Patent Application
20050026610
The invention relates generally to wireless networks, and more particularly to wireless network configuration and power level adjustment for network performance optimization.
The proliferation of laptop and hand-held portable computers has produced a concomitant need for robust, reliable, and high performance wireless networks to maximize the mobility advantages of these devices and increase the ease of construction and management of these wireless networks. Current wireless networks, such as IEEE 802.11b, 802.11a, 802.11g, (etc) networks, are subject to certain limitations that can limit a mobile user's network performance and reliability. For instance, only a very limited number of radio channels are available. In the current state of the art, wireless access points cannot effectively share the same channel in the same area because of radio and control protocol interference. So, bandwidth over a given area is limited by the number of non-overlapping channels available. Also, current wireless networks require manual site engineering to control the placement of access points and channel distribution between access points, raising the cost and complexity of the wireless network installation process. Furthermore, user roaming between wireless access points is inconsistent. Once associated with an access point, a user will tend to remain associated with that access point even if another access point is capable of providing higher performance for the user. It would be desirable to provide wireless networking solutions which overcome the above described inadequacies and shortcomings of current wireless networks.
In accordance with the principles of the invention, various apparatus, methods, and computer program products are provided to improve the performance and ease of management of wireless communications environments. For example, a mechanism is provided to enable access points (APs) to perform automatic channel selection. A wireless network can therefore include multiple APs, each of which will automatically choose a channel such that channel usage is optimized. Furthermore, APs can perform automatic power adjustment so that multiple APs can operate on the same channel while minimizing interference with each other. Further aspects of the invention are used to cause load balancing of stations across APs so that user bandwidth is optimized. Novel movement detection schemes provide seamless roaming of stations between APs. These and further aspects of the invention enable the provision of automatically configurable, high performance wireless communications environments.
In accordance with the present invention, a fully automatic control system is provided for wireless communications environments. Referring to
Various types of wireless communications environments 10 exist. Wireless communications environments include for example wireless data networks and wireless I/O channels. An example of a wireless data network is described in “IEEE Standard for Information technology—Telecommunications and information exchange between systems—Local and metropolitan area networks—Specific requirements—Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications—Amendment 1: High-speed Physical Layer in the 5 GHz band”, incorporated herein by reference (hereinafter “802.11”). Furthermore, various different 802.11 “modes” are defined. For example, in IEEE 802.11 compatible wireless networks, wireless devices may be arranged in an “infrastructure mode”, whereby the network is configured such that STAs 16 communicate with other network devices via an AP 12, as shown in
According to aspects of the invention, an arbitrary number of wireless access points (APs) can be placed in arbitrary positions, and all APs and STAs will automatically configure themselves for optimal channel usage, power levels, and STA/AP associations. So, in a wireless networking environment, channel usage is optimized while interference between APs is minimized. Wireless devices such as wireless enabled laptops or hand-held computing devices or Internet protocol telephones, are transparently and seamlessly distributed between APs such that network performance is optimized from the perspective of the user of the wireless device. And, in a wireless I/O channel environment that might be employed for example in a home, audio, video, and other appliances may be moved without performance degradation, and channel usage for each appliance may be optimized so that the appliances do not interfere with each other.
In order to expedite the understanding of the invention, certain examples will be described as they apply to the relatively well known 802.11 wireless LAN architecture, with the understanding that the principles of the invention apply more generally to any wireless communications environment. A preferred implementation of the inventive principles will then be described as embodied in an 802.11 wireless network
The following aspects of the invention contribute to its advantages, and each will be described in detail below.
Since exemplary examples will refer to the 802.11 networking environment, the following information provides relevant context, while understandably not limiting the invention to 802.11 environments.
In an 802.11 network, APs periodically send frames called “Beacons”. STAs listen for Beacons. When an unassociated STA (i.e. a STA that is not yet able to communicate on the wireless network) hears Beacons at what it deems to be a reasonable power level, it can attempt to authenticate with the AP sending the Beacons, and then associate with that AP. Once authenticated and associated, the STA is able to send data frames to other STAs on the wireless network via the AP.
More particularly, APs and STAs send and respond to three different types of frames, known as Class 1 Frames, Class 2 Frames, and Class 3 Frames. Class 1 Frames include control frames and management frames, and can be sent regardless of whether a STA is authenticated and associated with an AP. A Beacon is a type of Class 1 frame. Class 2 Frames are sent only once a STA is authenticated, and include for example association request/response messages. Class 3 Frames can be sent only if associated, and include data frames.
In order to maximize user bandwidth and throughput, the invention automatically optimizes the operation of multiple APs for a given wireless communications environment. In accordance with the exemplary IEEE 802.11a networking standard, 13 non-overlapping frequencies are available for use by the APs. Each AP is capable of transmitting and receiving data at a maximum rate of 54 Mbps. The actual rate at which data is transmitted and received between an AP and a STA depends upon many factors, including the distance between the AP and the STA, the structures located between the AP and the STA, and the environmental interference occurring on the particular frequency. One skilled in the art will realize that the invention is not limited by the maximum data rates of current wireless technology, nor is it limited by currently understood radio frequency attenuation factors. The principles of the invention will continue to be applicable as wireless technology evolves.
Consider an area such as the wireless network shown in
Now, referring to
In accordance with the invention, a Dynamic Radio Control Protocol (DRCP) provides a mechanism for an arbitrary collection of STAs and APs to automatically control the frequency and power of their radios in order to extend the properties exemplified in
DRCP Preclaim and claim messages are exchanged between APs during AP initialization, and are used to aid automatic channel selection in accordance with the invention. DRCP Announce messages are sent by APs and received by STAs during STA optimization. These Announce messages inform invention-enabled STAs of available invention-enabled APs to which they may choose to associate, and provide information about APs that STAs can use to aid a decision as to whether to request to roam to another AP. DRCP Bid messages are sent by STAs to APs during STA optimization. These messages inform invention-enabled APs of invention-enabled STAs that are requesting association to the APs. DRCP Accept messages are sent by APs to STAs in response to DRCP Bid messages. These messages inform a STA that it may associate with the AP it is requesting to associate with. DRCP registration request and acknowledge messages are exchanged by invention-enabled APs and STAs to indicate to each that the other is DRCP capable.
The DRCP protocol employing these messages will first be described as used in a generic wireless communications environment. The detailed implementation of each of these messages will then be further described in terms of a preferred embodiment in an 802.11 environment.
The aspects of the invention are now described as they apply to AP initialization and optimization, and then as they apply to STA initialization and optimization. It is noted that, though many of the inventive aspects described herein with regard to STAs and APs are advantageous when implemented, they are not required to be implemented in a wireless communications environment. Performance advantages are achieved when only the APs, or only the STAs, or both implement one or more of the various aspects of the invention.
1. AP Initialization
During AP initialization, APs perform automatic channel selection. In accordance with the channel selection aspect of the invention, APs located in the same wireless network automatically select channels for operation such that they do not interfere with nearby APs. The invention contemplates that different bands of frequencies are available, for example based on 802.11 version and the country in which the network is deployed. According to a preferred embodiment, APs attempt to select a channel, in each band in which the AP is equipped to operate, which is least likely to interfere with other APs that are already deployed. APs also quarantine channels in accordance with rules associated with regulatory domains (Europe, etc.) so they don't interfere with other wireless applications (radar, etc.). In the event that one AP selects a free channel, and another AP selects the same free channel at the same time (i.e. a channel selection “Collision”), the APs' media access control (MAC) addresses are used as a tie breaker. If the other AP is a standard AP that does not include the improvements associated with the current invention, then the invention-enabled AP will direct its own radio to the “next best” channel. The AP repeats the channel selection phase for each band of frequencies.
More particularly, referring to
For each band:
Scan Intervals occur periodically. During a Scan Interval (step 20), the AP 12 passively scans all channels which the AP supports within the band (step 22). The AP 12 gathers a list of active APs 12, the channels on which they are operating, and the power at which the beacons from each AP 12 was heard. This information is used to build a table called a channel map 24 (step 26), which contains a list of all APs 12 heard from, the channel on which they were heard, and the signal strength at which they were heard. There is a separate channel map 24 for each band. The AP 12 sorts the channel map to produce a list of APs 12 in ascending order of power level (step 28).
Referring to
If the AP 12 selected a channel having the weakest signal strength, the AP 12 notes the channel-ID of the channel that it has selected, the received power level on the channel, and the AP-ID of the AP that generated that power level (step 40). It will use the power level value as a baseline against which to detect increases in received power on its selected channel. If the AP 12 selected an empty channel, the baseline power level will be the AP's noise floor.
The AP 12 then advertises its intention to use the selected channel by periodically transmitting DRCP claim messages during the claiming period (step 42). Claim messages are transmitted at full power. During this claiming period, the AP 12 receives all Beacons, DRCP claim messages, and DRCP Announce messages transmitted on the currently chosen channel (step 44) and uses the information contained therein to build an “Other APs” table 46 (
If another AP 12 starts to radiate significant energy on the selected channel, one of two events must have occurred. The new AP 12 is either not running DRCP, or a conflict has occurred with another DRCP-active AP, where a race condition has caused the other DRCP-active AP to select the same channel at the same time. This is called a Channel Selection Collision (CSC).
At the end of the claim period (step 50), the AP 12 stops sending claim messages and evaluates the information it has collected, its CSC data, to determine if a CSC has occurred. It looks to see if the received power in any entry is greater than the baseline power level it recorded for the channel (step 52). If so, it looks to see if the received power is exceeded in at least half of the power level values for the entry (step 54). If so, the AP 12 checks to see whether the AP in the entry is DRCP capable (step 56).
If the other AP is not DRCP active, the AP 12 defers to the non-DRCP-active AP and starts the entire channel selection process over again.
If the other AP is DRCP-active, then a CSC is assumed to have occurred. When a CSC has occurred, the MAC address of the other AP is compared to the MAC Address of this AP 12. If the MAC address of this AP 12 is numerically higher than the observed MAC address (step 58), this AP 12 starts the channel selection process over again.
If at the end of the claiming period, the AP has succeeded in claiming the selected channel, it begins running on the channel. The AP starts beaconing, begins sending DRCP Announce messages, and prepares to enter the Optimization stage in order to run its Auction and Power Adjustment functions (step 60).
A variation on the channel selection process of
Referring to
Scan Intervals occur periodically. During a Scan Interval (step 100), the AP 12 scans all channels which the AP supports within the band, receiving Beacons and Announce messages (step 102). The information received in the Beacon and Announce messages is used to build a table called the “Scan table” (step 104). An example of a Scan table 106 is shown in
As the scan progresses, the rxPowerSampleCount values in the Scan table 106 are monitored, and if any one of them exceeds a threshold, herein labeled “threshold 1” (step 108), the AP 12 proceeds to step 110 to build a channel map. In addition, if any entryAge value in the table exceeds a certain threshold “threshold2” (step 112), the AP 12 will proceed to step 110. Also, if the number of scans completed exceeds a certain threshold “threshold3” (step 114), the AP 12 will proceed to step 110. Otherwise the AP 12 continues scanning and updating the Scan table 106.
An example of a preferred channel map 116 is shown in
RxPowerAvg[I]=Scan table[i].rxPowerRunningTotal/Scan table[i]rxPowerSamplecount;
For each channel, the AP-ID with the highest rxPowerAvg value is entered in the channel map 116. The AP's rxPowerAvg value is entered into the channel map 116 as the highestPwrlevel parameter.
In certain network implementations, it is undesirable to locate two operating APs 12 within a certain distance of each other, because doing so does not increase network performance and may reduce it. So, according to a preferred option, once the channel map has been assembled, the AP 12 checks the channel map to see if any of the highestPwrlevel values in the map exceed a certain threshold power level (step 118). The threshold power level is chosen to indicate that the AP 12 is located too close to another AP. If any highestPwrlevel exceeds the threshold power level, the AP 12 is placed in “Standby mode” (step 120). The AP 12 in Standby mode waits for a period of time herein referred to as “Standby Interval” before returning to start another scan interval (step 122).
If no highestPwrlevel values are found to exceed the threshold power level, the AP proceeds to assemble a triplet channel map 124 (step 126). As shown in
The AP 12 then records the baseline power for the selected channel as the highestPwrlevel value for the channel, and records the AP-ID of the AP at that power level (step 130). A Preclaim Interval is now entered (step 132). During the Preclaim interval, the AP transmits Preclaim messages on the selected channel (step 134). The AP 12 also receives Beacon, Announce, and Preclaim messages on the channel (step 136). The AP 12 uses the received messages to update the Scan table (step 138). The AP 12 continues to update the Scan table until one of two Preclaim intervals expires. During the min Preclaim interval (step 140), the rxPowerSampleCount value for each AP-ID on the selected channel is checked to see if a minimum sample size threshold has been exceeded. If so, the Preclaiming interval ends and the AP proceeds to check to see whether too many APs are operating on the selected channel (step 142). Otherwise, the Preclaiming period extends until the end of the max Preclaim interval (step 144).
In some network environments, and in particular environments of limited area, the operation of too many APs on the same channel does not increase network performance and may cause a performance reduction. So, according to a preferred option, the AP checks to see if there are too many APs operating on the network (step 142). The number of different AP-IDs present in the Preclaim APs table is used to make this determination. If there are too many APs operating on the selected channel at a power level greater than a defined threshold, the AP 12 enters Standby mode
If there are not too many APs operating on the selected channel, the AP 12 calculates an “adjacency vector sum” (step 146). The adjacency vector sum is calculated over all APs on all channels as:
Adjacency vector sum=sum(Preclaim APs[i].ReceivedPowerTotal/Preclaim APs[i].count)
The adjacency vector sum is used as a tiebreaker if necessary during the claiming interval, to be further described.
The AP 12 now enters the claiming interval (step 148). During the claiming interval, the AP 12 transmits claim messages on the selected channel (step 150). Claim messages include the adjacency vector sum. The AP 12 also receives all Beacon, Announce, and claim messages on the selected channel (step 152). The AP 12 uses the information contained in the received messages to build a “claim APs table” 154 (step 156). An example of the claim APs table is shown in
The AP 12 then evaluates the claim APs table 154 (step 160) to ascertain whether any other APs were heard from. If no entries exist in the claim APs table 154 (step 162), the AP 12 has “won” the selected channel. The AP 12 can then begin Beaconing and sending Announce messages on the channel (step 164). If the claim APs table 154 includes one or more entries, then if the selected channel was empty at the beginning of the claiming Interval (step 166), the AP 12 concedes the channel and returns to re-scan (step 100). Otherwise, if the channel was not empty, then if the ReceivedPowerlevel values for all entries in the table are less than the baseline level recorded during Preclaiming plus a threshold level (e.g. 2 db) (step 168), then the AP has won the channel (step 164). However, if the AP does find a ReceivedPowerlevel value that exceeds the baseline power level plus the threshold level, and that entry is associated with an AP-ID that was not the one recorded on the channel during the Preclaiming Interval, then the AP checks the entry's DRCP flag (step 170). If the DRCP flag indicates that the AP-ID associated with the entry is not DRCP capable, then the AP returns back to the scan interval to re-start the channel selection process. If the DRCP flag indicates that the AP-ID associated with the entry is DRCP capable, then the AP compares its adjacency vector with the adjacency vector received in the other AP's claim messages (step 172). If the AP's adjacency vector is less than the other AP's adjacency vector, the AP concedes the channel and returns to re-scan. If the AP's adjacency vector is not less than the other AP's adjacency vector, then the AP checks to see if the adjacency vectors are equal (step 174). If they are, the MAC addresses of the two APs are compared (step 176). If the AP's MAC address is greater than the other AP's MAC address, the AP has “won” the channel (step 164). Otherwise, the AP concedes the channel and returns to re-scan (step 100). One skilled in the art will realize that the MAC address comparison decision is arbitrary and can be made in the opposite manner.
1. AP Optimization
Once an AP is running on a channel, it continuously performs the following functions to optimize its configuration in the wireless LAN:
In order to perform Radio Power Adjustment, the AP 12 maintains a number of tables. The tables include information received from other APs and STAs operating on the selected channel. This information is used to ascertain such things as power levels of other devices on the channel, and distances between devices on the channel, in order to control the AP's power level.
2.a.1.1 AP KnownAPs Table
Each AP 12 maintains a number of tables that it consults to perform power adjustment. One such table is the AP KnownAPs Table 200. As shown in
Referring to
The AP-ID, TP Backoff, and Max Power fields are extracted from each DRCP Announce and Hello message received. The TP Backoff value is stored as a list of values for each message received from each AP-ID. The sample size is the number of TP Backoff values received for each AP-ID.
Since Announce messages are only sent by DRCP-enabled APs, the AP 12 also marks the entry as DRCP-active. APs sending Beacons which contain no DRCP fields are not marked DRCP-active. For each message received, the AP adds the TP Backoff value to the received power level to determine a normalized received power level, as follows:
normalized—power=avg(received—power+tp—backoff);
Accounting for the TP backoff in the normalized power level provides a value that is consistent and can be used for comparison with power level measurements from other APs.
Received beacons do not explicitly carry a TP backoff value, however, since beacons are always transmitted at full power they effectively carry a TP backoff value of zero. Thus, the AP 12 can update the AP KnownAPs table 200 based on a received Beacon. In this case the AP 12 stores the TP backoff value as zero, and sets the normalized power and the Max Power to the received power level value.
Additionally, the AP 12 keeps an Age for each entry. The Age is reset to zero, “0”, each time a Beacon or DRCP Announce is received from the AP corresponding to the entry. Entries are aged as part of the AP Power Adjustment process.
2.a.1.2 AP AssociatedSTAs Table
APs 12 also continuously maintain a table of the STAs 16 that are associated with it—the AP AssociatedSTA table. Referring to
As shown in
During the channel selection process, the AP 12 transmits at maximum power, that is, it uses a TP backoff value of zero. Once the AP 12 has successfully claimed a channel, it calculates a TP Backoff value and adjust its transmit power for data transmissions down, in accordance with this value to minimize same channel interference and maximize channel/bandwidth re-use. The calculation of the TP Backoff value is now further described.
Generally, with reference to
Once the Max TP backoff is calculated, the AP then scans the AssociatedSTA table to ascertain the distance to the farthest associated STA 16 (step 264). The distance to the farthest associated STA 16 is compared to the distance to the closest AP 12 operating on the same channel (step 266). If the distance to the farthest associated STA 16 is less than the distance to the closest AP 12, the AP's TP Backoff value is left at Max TP Backoff (step 268). If the distance to the farthest associated STA is greater than the distance to the closest AP operating on the same channel (step 266), then the Max TP backoff value is adjusted back down to accommodate this STA (step 270), and the AP's TP backoff is set to this adjusted value (step 272). This power adjustment is periodically repeated to account for changes in the AP KnownAPs table and the AP AssociatedSTAs tables change (step 274). Power adjustement may be repeated every second, for example, in an 802.11 environment.
In accordance with the preferred embodiment for use in a wireless data networking environment, as shown in
If the Avoid Other WLANs flag is not set, the AP 12 sets its TP Backoff to 0 (step 282). The AP 12 will therefore transmit at full power as long as the Avoid Other WLANs flag is not set. If the Avoid Other WLANs flag is set, the AP proceeds with the power adjustment process and will not interfere with other APs even if they are operating on a different network.
The AP 12 processes the information in the AP KnownAPs table every Hello Interval (step 284), to perform power adjustment. The Hello Interval is architecturally and design dependent. In an 802.11 environment, the Hello Interval may be for example the 100 ms beacon interval. The knownAPs table entries are first aged before other processing is performed. The age of each entry is incremented (step 286), and any entries whose age exceeds Max AP Entry Age are deleted (steps 288, 290). Thus, if an AP hasn't been heard from in a while, it is aged out of the table to prevent it from affecting this AP 12's power adjustment calculations.
As previously described, the normalized power field of an entry, “n”, of the AP KnownAPs table is determined by averaging over several received power measurement samples. The greater the number of samples used to derive the value, the more accurate the measurement. The AP accounts for the inaccuracy in this value by subtracting a standard error, based on the sample size, from the normalized power level, as shown in the following formula. Note, in an 802.11 environment, the normalized power level is expected be a negative number, ranging in value from −25 to −90 dBm.
AP knownAPs[n].corrected—power=AP knownAPs[n].normalized—power−getStandardError(sample—size);
In this formula, the function “getStandardError” would return the standard error for the sample size of “sample_size” in each entry “n”. For example, table I in
The AP KnownAPs table is then scanned for the entry corresponding to the AP which was heard at the highest corrected_power level. This corrected_power level is compared to the AP 12's noise floor (step 294). (The AP's noise floor is a measure of background power on the channel.) If the AP 12 finds that the highest corrected_power level is less than the AP 12's noise floor, then the AP 12 may transmit at maximum power without interfering with the other APs on the channel. It therefore leaves its Max TP Backoff value at 0 (step 296). If the AP 12, however, finds that the highest corrected_power level is higher than the AP 12's noise floor, it needs to set a TP Backoff to avoid interfering with that AP. In this case the AP 12 calculates Max TP Backoff by subtracting its noise floor from the corrected_power associated with that AP (step 298).
Once it has calculated Max TP Backoff, the AP 12 must then determine if there are any associated STAs 16 that are farther away (ie who's signal strength is weaker than) the highest_power_AP. The AP 12's AssociatedSTAs table is analyzed to find the STA 16 that is the greatest distance from the AP (step 300). The normalized power and sample size values for this STA are used to calculate a corrected_Power value for this STA as previously described. The lowest power STA's corrected_power level is compared to the AP 12's noise floor (step 302). If the corrected_Power level is less than the noise floor, then the AP 12 needs to run at full power to cover the STA, so the AP TP Backoff value is set to 0 (step 304). If the corrected_power level is greater than the AP 12's noise floor, then the STA TP Backoff value is set to the corrected_power level minus the noise floor (step 306).
Next, Max TP Backoff is compared to STA TP Backoff (step 308). The AP's TP Backoff (“my TP Backoff”) is set to the lower of the two backoff values to avoid interference with the closest AP while ensuring coverage for the farthest STA. So, If STA TP Backoff is less than Max TP Backoff, then my TP Backoff is set to the Max TP Backoff value (step 310). If STA TP Backoff is greater than Max TP Backoff, then the my TP Backoff value is set to the STA TP Backoff value plus some minimum signal-to-noise ratio (step 312).
The AP 12 then adjusts its transmit power by the value of my TP Backoff (step 314), and uses my TP Backoff as the value of TP Backoff in its Announce messages. According to the preferred embodiment, the AP will transmit data at a power level adjusted by TP Backoff, but will tranmsit DRCP management messages (e.g. Claim, Announce, Accept) at full power. So, APs can always hear management messages passed between each other.
Furthermore, various different wireless networking architectures may provide a mechanism for clearing the wireless channel, further increasing the probability that the management messages will be received. For example, in the 802.11 architecture, the AP issues a Clear to Send (CTS) message to clear the channel (step 316), and then sends a DRCP Announce message at maximum power (step 318). After sending the Announce message at maximum power level, the AP resumes use of its calculated TP backoff value for data packets to minimize same-channel interference, as described above.
2.a.1.3.i AP Power Adjustment During Station Movement
When a STA 16 is associated with an AP 12, the AP 12 keeps track of the distance between the STA 16 and the AP 12. If the AP 12 is using a non-zero TP Backoff value, and the AP 12 ascertains that the STA 16 is moving out of range of the AP 12 at backed off power, then the AP 12 can adjust its TP Backoff value to accommodate the STA 16's movement.
The distance between an AP 12 and an associated STA 16 is calculated and stored in the AssociatedSTAs table in units of “Banzais”. The Banzai is a unit of distance derived from a measurement of received signal strength from an AP 12 operating with a known transmit power backoff. In an 802.11 environment, for example, a received signal strength measurement is generally expected to range in value from −25 dBm to −90 dBm, but depending upon possible antenna gain at the high end or sensitivity at the low end, may range from 0 dBm down to −100 dBm. A transmit power backoff is generally expected to range in value from 0 dB to 65 dB. Given a received signal strength measurement of “received_power” and transmit power backoff of “TP Backoff” from an AP, the distance to that AP in Banzais is calculated as follows:
distance—in—banzais=ABS[MIN[0, (received—power+tpbackoff)]]
Algorithms for movement detection are described in more detail later. For purposes of AP power adjustment, as previously described, the AP 12 collects multiple samples of received_power and TP Backoff values for each STA 16 over time, and the distance between the AP and the STA is calculated over all these samples. The AP 12 detects movement by continuously checking to see if the difference between the distance to each station, derived from a set of long term samples, is sufficiently smaller that the current distance measurement based on the most recently collected samples, to indicate that the STA 16 is moving away from its AP 12.
If the AP 12 detects that the STA 16 is moving away and the STA 16 is within a given Short Term Standard Error Banzais of the current edge of the transmit signal (based on the current TP Backoff), then the AP 12 switches to a TP Backoff of 0 until it no longer has any moving STAs associated with it.
More particularly, referring to
If the AP 12 detects that the STA 16 is moving, the AP 12 sets its TP Backoff to 0 so that it transmits data at maximum power (step 328). The AP 12 remains at maximum power until it detects that the STA 16 is no longer moving, or quiet. If the AP 12 determines that it is no longer moving before the STA 16 loses the association to its AP, the AP 12 resumes normal processing of received signal strength to determine a new appropriate TP Backoff value as previously described (step 334).
Once the AP 12 detects that the STA 16 is moving, it begins using another test to detect when the STA 16 has stopped moving. To detect that the STA 16 has stopped moving, the AP 12 compares the distance to the STA 16, derived using the Long Term Sample Size, to the distance derived from the most recent Short Term Sample Size samples. The AP 12 looks to see when this difference is less than just the standard error in the two measurements to determine that the STA 16 has stopped moving. This test is performed as follows (step 330):
2.b AP Auction
The purpose of the Auction is to accomplish the distribution of STAs 16 across APs 12 in a manner that optimizes wireless communications performance. The goal is to have STAs 16 associate to their nearest AP 12 while taking loading (the sum of the individual loads of the STAs 16 already associated to the AP 12) into account. This allows the RF footprints of the APs 12 and STAs 16 to be minimized, while ensuring that no AP 12 is overloaded.
STAs 16 learn of available APs 12 through the Announce messages transmitted by the APs 12. As will be further described with regard to STA optimization, a STA 16 calculates a “biased distance” to each AP 12 it hears from, including its own AP, using the received power and loading information from the Announce messages. A STA 16 will send a Bid message to an AP that is “better” than the STA's current AP, where better means that the AP has a lower biased distance. The Bid message contains the value of the difference between the biased distance from the STA 16 to the destination AP 12 and the biased distance to the STA 16's current AP. This value is called the biased distance delta.
In particular, referring to
At the end of Auction Interval (step 352), the AP 12 processes the received bid information. (The Auction Interval in an exemplary 802.11 environment may be on the order of, for example, 7.5 seconds.) The age of all bid entries is incremented by one (step 354) and then any bid entry whose age is greater than Max Bid Age is deleted (step 356). The list is then sorted by biased_distance_delta value (step 358).
The AP 12 selects the bid entries with the highest biased distance delta values, up to acceptsPerAuction entries, and sends a DRCP Accept message to each of the STAs 16 corresponding to those entries (step 360). The IDs of each STA 16 being sent an Accept is put in a list of outstanding accepts (step 362), and a count of accepted STAs who have not yet associated and registered is noted as numAcceptsOut (step 364). At this point the next auction period begins.
In addition to receiving DRCP Bids, the AP 12 also receives an indication any time a STA 16 associates to the AP 12. Referring to
3. STA Initialization:
The purpose of the STA initialization phase is to find and associate to a suitable AP 12 to provide the STA 16 with access to the wireless LAN, and to prepare for the operation of the DRCP protocol and algorithms.
Referring to
First, the STA 16 scans for beacons on all channels across all supported bands (e.g., STAs supporting multiple bands such as 802.11a/b/g will scan all channels in each of the a, b, and g bands.) (step 382) The AP 12 that is received at the greatest signal strength is selected (step 384).
Preferably, where multiple bands are supported, the selection of an AP 12 also takes into account a preference for the higher bandwidth bands so that, for example as implemented in an 802.11 environment, an AP 12 on an 802.11a or 802.11g channel is given some preference to one operating on an 802.11b channel.
Once an AP 12 is chosen, the STA 16 authenticates and associates to that AP (step 386). Any security policies that control association are executed at this point.
During initialization, the STA 16 also enables the DRCP protocol so that the STA 16 will receive DRCP Announce messages (step 388). The STA 16 then proceeds to the STA optimization phase (step 390).
4. STA Optimization
Once it has made its initial association and has access to the wireless LAN, the STA continuously performs the following functions to optimize its configuration:
The process by which the STA 16 canvasses the other channels to determine whether to send a DRCP Bid message is now described in further detail. In order to monitor DRCP Announce messages, a STA 16 periodically tunes its radio to the channels other than the one to which it is currently associated. However, the STA 16 must remain associated to its current AP 12 so that it does not lose data packets. So, packets must be buffered during the time that the STA 16 is canvassing. Various wireless communications architectures may provide different means for packet buffering.
Generally, referring to
The 802.11 power save mode is intended for use by STAs 16 so that they can turn off their radios for periods of time in order to save power. STAs 16 can indicate to APs 12 that they are entering this power save mode. In response, APs 12 buffer the STAs' packets while the STAs 16 are “sleeping”. APs 12 periodically send special Beacon messages to the STAs 16. STAs 16 wake up in response to these special Beacon messages. These Beacon messages include information as to whether any data is buffered for the STA 16. The STA 16 “wakes up” if data is buffered for it.
STAs 16 operating in an 802.11 environment in accordance with the preferred embodiment of the invention use the 802.11 power save mode to go off-channel and canvass other channels for DRCP Announce messages. After the channel canvass is complete, the STA 16 reverts to normal power save mode. Stations that have not been set to Power Save Mode by management are caused by the STA 16 to act as if they've been set to the Power Save Mode. STAs 16 that have already been set to Power Save Mode by management will have even more time to canvass.
In particular, referring to
Stations that have not been set to Power Save Mode by management are caused by the STA 16 to act as if they've been set to Power Save Mode, with listen interval set to the minimum value (i.e., every Beacon). STAs 16 that have already been set to Power Save Mode by management will have the most amount of time to canvass.
During the canvass time the STA 16 tunes its radio to a different channel in order to passively listen for beacons and DRCP Announce messages. The STA 16 keeps track of which channels have been canvassed, stepping through all of the channels until all supported channels have been canvassed. The STA 16 keeps track of all DRCP Announce messages, and the power level at which they were received.
4.a.1 STA KnownAPs Table
The STA 16 receives beacons and DRCP Announce messages from all APs 12 within the range of its radio. These are processed to build a table of all known APs 12, the “STA knownAPs” table 430, as shown in
The STA Known APs table 430 is built as shown in
The STA 16 stores the following fields from received beacons in the STA knownAPs table entries:
The STA 16 stores the following fields from received Announce messages in the corresponding STA knownAPs table entry:
The STA 16 also notes the received power level that accompanied the beacons and Announce messages and uses these values along with the TP backoff values to calculate the distance to the APs in Banzais, as previously described. (Again, since non-DRCP APs will always send beacons at full-power, the TP Backoff value for these is set to 0.)
The Received Power and TP Backoff entries are lists, where each entry in each list corresponds to a Beacon or Announce message received for the corresponding AP-ID. The received power level value and correspondingly, the distance in Banzais, are subject to variability in the RF channel. The STA 16 saves a number of these distance measurements for each entry in the knownAPs table, so that it can use averaging to compensate for this variance, to be further described. For its own AP (i.e. the AP to which the STA is currently associated), the STA 16 averages over a relatively large number of distance values, herein referred to as “Long Term Sample Size” distance values. For all other entries, the STA uses fewer, “Bid Sample Size”, distance values.
Additionally, the STA keeps an Age for each entry. The Age is reset to zero, “0”, each time an Announce message is received from the AP corresponding to the entry. Entries are aged as part of the STA Bidding process, described below.
4.b STA Power Adjustment
Referring to
4.c STA-Bidding
Each time the STA Canvassing function completes a canvass of all channels, the STA 16 analyzes the information in the STA knownAPs table to see if there is a potential “better” AP 12 with which to associate. The notion of what constitutes a better AP takes into account the distance to the AP in Banzais, the available data rate, and the loading (number of associated STAs) on the AP, if known.
Referring to
As mentioned, the STA 16 uses averaging to compensate for variance in its distance measurements. It requires Long Term Sample Size distance values for the STA knownAPs entry corresponding to its AP, before it performs further processing of the table (step 456). Once the STA 16 has Long Term Sample Size distance values for its own AP, it then waits until it has Bid Sample Size distance values for all entries in the knownAPs list at that time before it begins looking for a better AP (step 458). This is to avoid making a decision to move to a new AP before it has sufficient information about the other APs in the network. However, to avoid the potential for waiting indefinitely, it will not delay processing the knownAPs list for any new APs that were added after it has Long Term Sample Size distance values for its own AP.
Working with the entries for which there are sufficient distance measurement samples, the STA 16 looks for a potential better AP. In summary, a biased distance is calculated for each entry, which takes into account the available data rate as well as the loads on the APs (step 460). The data rate is deduced based on the received signal strength and the technology being used (i.e., in an 802.11 environment, the 802.11 mode of operation (a,b,g)). After calculating the biased distances for all of the entries in the STA knownAP table, the AP with the lowest biased distance is considered to be the best candidate and, if it appears better than its current AP (step 462), a Bid is sent (step 464).
More particularly, referring to
As previously described, the distance field in the STA knownAPs table 430, knownAPs[n].distance per entry, is the distance in Banzais, to the corresponding AP, averaged over a number, Bid Sample Size, of measurement samples. As previously noted, this value is subject to variance in the RF channel. The bid selection process should preferably yield the choice of a “better” AP only when the new AP actually would provide better performance for the STA, and not when the new AP simply appears to be better due to this variance. The variance in the new AP's distance measurement is represented by a “Bid Sample Std Error” value related to the bid sample size, and the variance in the current AP's distance measurement is represented by a “Long Term Std Error” value related to the long term distance sample size. It is advantageous to minimize the error in the distance measurement for a given entry n, in comparison to the STA's current AP. This is done by using a corrected distance that is set to the distance to the STA's current AP, if the entry falls within the sum of the standard errors on the two distance measurements. The corrected distance, “corrected_distance”, is determined as follows:
4.c.1 Distance to Load Factor Conversion
The corrected distance to each AP 12 is recorded in the STA knownAPs list. This distance is then used in conjunction with data related to the particular wireless environment in which the AP 12 is operating to derive an estimate on the expected load factor for the STA. For example, in an 802.11 environment, the distance and 802.11 mode (a,b,g) are used to retrieve the expected data rate for the STA 16 from the distance_to_rate table:
An example of distance to rate calculations for 802.11 modes is shown in Table II in
Then, the expected load for this datarate is retrieved from the rate_to_load table:
An example of a rate_to_load table for 802.11 networks is shown in Table III in
Any entries in the knownAPs list that represent non-DRCP APs need to be given default values for their load_factors so that they can be considered by the Stations for associations as well. These default load_factors are derived from a default number of STAs per AP value that should be consistent across the network and a default “average data rate” per technology. That is:
When determining the load of the STA 16's current AP (myAP), when myAP is a non-DRCP AP, then its default load_factor value is preferably incremented by the STA's load on that AP. This helps to support a consistent view of the load of that AP both before and after a STA associates with it—that is, since the STA adds its own load to the default load of its prospective (non-DRCP) AP before it makes a decision to associate with it, it must also add its load to the default load for this AP after it has associated with it.
4.c.2 Biased Distance Calculation
Using the my_load_factor to the AP 12, the load_factor currently on the AP 12 (received from Announce messages) and the corrected distance to the AP 12, the STA 16 calculates a biased_distance value to account for the loading on the prospective AP 12 in comparison to the loading on the STA 16's current AP, as shown in
Next, a biased distance to the STA 16's current AP 12 is calculated to account for the loading on the current AP 12 relative to the loading on the prospective AP. This calculation is made as follows:
Finally, the difference between the biased distance to the prospective AP and the relative, biased distance to the STA 16's current AP is determined, as follows:
After calculating the biased_distances and biased_distance_deltas for all of the APs in the knownAPs list, the STA 16 checks to see if any of the biased_distance_delta values are positive (step 496). If not, then the STA 16's current AP is still the best AP, so the STA 16 remains associated with its current AP (step 498). Of any positive biased_distance_delta values, the best AP is the one with the highest positive biased_distance_delta value (step 500). If the best AP is not DRCP enabled (step 502), then the STA 16 associates with that AP (step 504). If the best AP is a DRCP AP (step 502), then a Bid is sent (step 506) and the STA 16 resumes normal operation until it either receives a DRCP Accept or it completes another Canvass Sample Number passes of all channels. If there is more than one AP with the same highest biased_distance_delta values (step 508), the STA 16 checks to see if any of them is the last AP to which it Bid (step 510) and if so, it selects that one again (step 512).
If a DRCP Accept is received with the AP-ID matching the selected APs AP-ID (step 514), the STA sets its TP backoff value to zero (step 516) and associates with the AP from which the Accept was received (step 518). The STA 16 now sends a DRCP Registration Request to the AP (step 520) and starts a timer (step 522) to go off every Registration Timeout Interval. When the timer expires (step 524), the STA 16 sends out another DRCP Registration Request and resets the timer. Upon receipt of a DRCP Registration Acknowledge (step 526), the timer is disabled (step 528).
If no DRCP Accept is received (step 414) in response to the STA's Bid message after a certain period of time (step 530), the STA remains associated with its current AP.
After the STA 16 has associated with a new AP, the STA 16 waits until it has collected a large number, Long Term Sample Size, of distance measurements to its new AP before it resumes this process of evaluating the knownAPs table for bidding.
4.d STA Movement Detection
When a STA 16 is associated with an AP 12, the STA 16 receives DRCP Announce messages from all APs 12 within the range of its radio. Referring to
The STA 16 continuously collects Short Term Sample Size distance values to the current AP (step 542). To detect movement, the STA 16 compares the distance to its AP 12, derived from averaging over the long term using Long Term Sample Size, to the short term distance, derived from averaging the most recent Short Term Sample Size samples. This difference is compared to a Moving Threshold value plus the standard error in these two measurements in order to eliminate false movement detection. A station is considered to be moving when the short term distance exceeds the long term distance by more than the Moving Threshold plus the sum of the errors in the two measurements. The following pseudo code describes this comparison.
Once the STA 16 detects that it is moving (step 446), and as long as the STA 16 does not detect that it has stopped moving, the STA 16 refrains from participating in the bidding process (step 548), seeking a new AP only if warranted by the deterioration of its current association. If the STA 16 determines that it is no longer moving before the STA 16 loses the association to its AP, the STA 16 resumes normal operation including participation in the bidding process.
As in movement detection, to detect that the STA 16 has stopped moving, the STA 16 compares the distance to its AP 12, derived using the Long Term Sample Size, to the distance derived from the most recent Short Term Sample Size samples. The STA 16 looks to see when this difference is less than just the standard error in the two measurements to determine that the STA 16 has stopped moving. The following pseudo code describes this test.
5. Software Architecture
In accordance with the preferred embodiment, the previously described functionality is implemented in software in APs 12 and STAs 16 respectively. Referring to
Referring to
Referring to
These DRCP messages are now described in further detail. Generally, DRCP messages could be encoded as standard LLC data frames, and the invention does not preclude such an implementation. But, according to the preferred embodiment as implemented in an 802.11 networking environment, DRCP management messages are encoded as new types in existing Class 1 Frames. DRCP messages are addressed either to a Group MAC Address, or to an individual MAC address, and are distinguished by the presence of the DRCP Protocol Identifier in the Protocol Identification Field of a SNAP PDU.
The DRCP messages are now described in detail as they operate on an IP WLAN. Some DRCP messages are transmitted as IEEE 802.11 MAC management frames of subtype Beacon on the wireless LAN only, while others are transmitted as data frames encoded as LLC 1 Unnumbered SNAP PDUs on the wireless LAN or the wired/wireless network between APs.
In accordance with the preferred embodiment, the SRMAs and ARMAs interact with the SRMs and ARMs to generate and/or collect information needed to produce or interpret DRCP protocol messages. It is noted that the DRCP protocol could be implemented over non-802-11 primitives without departing from the principles of the invention. The following describes the primitives used in an 802.11 environment.
5.a Enhancements to Standard 802.11 MAC Service Interface
The ARMAs and SRMAs transmit and receive DRCP messages over a standard 802 MAC Service Interface, with some enhancement. The receive interface is enhanced in both the STA and the AP to allow the SRM and the ARM to indicate to the SRMA and ARMA respectively, the power at which DRCP messages are received. In particular, the semantics of the 802.11 MA-UNITDATA.indication service primitive are modified as shown by the underlined text, to add a received power parameter as follows:
MA-UNITDATA.indication
The received power parameter specifies the signal strength, expressed in dBm, at which the MSDU was received. The received power value indicates the current level at which the sending device is heard, but does not provide an indication of whether or not the sending device is transmitting at full power. The potential power level at which a device might be heard can be determined when the transmit power backoff (i.e., the amount, in dB, by which the radio is turned down) in use by the device is also known.
5.b Enhancements to the Standard 802.11 Management Interface
BSS Description
The BSSDescription Parameter contains a list of elements that describe the BSS. An additional element is added to this list:
Send DRCP
This mechanism is provided to allow the ARMA to send DRCP Messages encoded in 802.11 Management frames of type Beacon.
MLME-SENDDRCP.request
This primitive is used by the ARMA to request that the ARM send a DRCP Message over the wireless media, encoded in an 802.11 Management frame of type Beacon. As shown in
The primitive parameters are as follows:
This primitive is generated by an ARMA to request that a DRCP Message be sent on the wireless media encoded in an 802.11 Management frame of type Beacon. DRCP claim and Hello messages are sent in this manner. As previously described, the ARMA may optionally quiet the channel before sending a DRCP Hello message by first by sending a CTS frame. In particular, if the Quiet Channel parameter is TRUE, the ARM transmits a Clear To Send (CTS) frame immediately prior to the Beacon transfer. The DRCP Hello CTS Destination MAC Address is placed in the Receiver Address (RA) of the CTS frame. The duration field of the CTS is set to the value of the CTS Duration parameter.
The fixed portion of the Beacon frame is as defined in the 802.11 standard. The DA is set to the Destination Address parameter value, the SA is the AP's MAC address, and the BSS ID is the DRCP Default BSS-ID. The variable portion of the frame is replaced by a single information element with an Element ID of DRCP Protocol, with a Length field value of the Message Length parameter and the Information field containing the DRCP Message.
MLME-SENDDRCP.confirm
This primitive confirms the transmission of a DRCP message to the ARMA. The primitive parameters are as follows:
This primitive is generated by the MLME as a result of an MLME-SENDDRCP.request to send a DRCP message encoded in an 802.11 Management frame of type Beacon. The ARMA is thus notified of the result of the Send DRCP request.
Power Management Fib
As previously described, one way that a STA can support periodic canvassing is to indicate to the AP that it is in power save mode, thereby causing the AP to buffer the STAs packets while the STA is canvassing. This mechanism supports a STA's ability to indicate to the AP that it is in power save mode, without actually going into power save mode.
MLME-POWERMGTFIB.request
This primitive requests the SME to use the power save mode interaction with the AP to allow time to canvass other channels. The primitive parameters are as follows:
This primitive is generated by an SRMA to cause the MLME to borrow part of the doze time (if the STA is in power save mode) or all of the doze time (if the STA is in active mode) in order to canvass other channels.
This request causes the SRM to:
This request prepares the SRM to:
This primitive confirms the change in power management mode to the SRMA. The primitive parameters are as follows:
This primitive is generated by the MLME as a result of an MLME-POWERMGTFIB.request to mimic power save mode. The SRMA is thus notified of the change of power mode indicated.
Power Save Start
This mechanism notifies the SRMA that it can begin to canvass.
MLME-PSSTART.indication
This primitive indicates to the SRMA the start of the power save cycle. The STA does not actually power off its radio and enter the sleep state at this point, but preferably, it should not transmit outgoing frames after sending this indication until it receives an MLME-PWRMGMTFIBCONTINUE.request. The primitive parameters are as follows:
This primitive is generated by an SME to indicate the start of power save cycle. The SRMA is thereby notified of the start of the power save cycle.
Power Management Restore
This mechanism further supports a STA's ability to indicate to the AP that it is in power save mode, without actually going into power save mode.
MLME-PWRMGMTRESTORE.request
This primitive tells the MLME that it should restore the user-configured power save mode. This primitive allows the SRMA to tell the MLME that it no longer needs to lie to the AP about power save (that control over power save is passed back to the MLME). The primitive parameters are as follows:
This primitive is generated when the canvass mechanism is taken out of service. The receipt of this primitive causes the SRM to restore the saved power management mode settings and:
This primitive confirms the change in power management mode to the SRMA. The primitive parameters are as follows:
This primitive is generated by the MLME to confirm that the SME has executed an MLME-PWRMGMTRESTORE.request. It is not generated until the change has been indicated. Upon receipt of this primitive, the SRMA is notified of the change of power mode indicated.
Power Management Fib Continue
Once canvassing is complete, this mechanism informs the SRMA that it “has control” of the radio and communicates power save state (awake or doze).
MLME-PWRMGMTFIBCONTINUE.request
This primitive tells the MLME that it's safe to enter the awake state and transmit frames if desired. The primitive parameters are as follows:
This primitive is generated when the SRMA is has completed canvassing. Upon receipt, the MLME enables transmission of user data frames, if necessary.
MLME-PWRMGMTFIBCONTINUE.confirm
This primitive confirms the change in allowed power management state. The primitive parameters are as follows:
This primitive is generated by the MLME to confirm that the SME has executed an MLME-PWRMGMTFIBCONTINUE.request. It is not generated until the change has been indicated. Receipt by the SRMA serves as notification of the change of the allowed power save mode.
Channel Out
This mechanism supports the ability to indicate to an ARMA that a channel has gone out of service.
MLME-CHANNELOUT.indication
This primitive reports to the ARMA that a channel that was previously available has become unavailable. The primitive parameters are as follows:
This primitive is generated by the MLME when a channel becomes unavailable. Receipt of this primitive causes the ARMA to remove the channel from its channel map.
Channel In
This mechanism provides the ability to indicate to an ARMA that a channel has gone into service.
MLME-CHANNELIN.indication
This primitive reports to the ARMA that a channel that was previously unavailable has become available. The primitive parameters are as follows:
This primitive is generated by the MLME when a channel becomes available. Receipt of this primitive causes the ARMA to add the channel to its channel map.
Beacon Notify
This mechanism supports the ability to detect any other APs using the same channel.
MLME-BEACONNOTIFY.request
This primitive requests the MLME to notify the ARMA whenever a beacon is received. There is one indication for each Beacon received. An indication is generated any time a Beacon is received on the current channel. The primitive parameters are as follows:
This primitive is generated by an ARMA when it wants to be notified of any beacons received on its own channel. Receipt of this primitive by an MLME causes the MLME to enable a mode whereby the ARMA will be notified if any Beacon is received.
MLME-BEACONNOTIFY.confirm
This primitive confirms the change in the beacon notification mechanism. The primitive parameters are as follows:
This primitive is generated by the MLME as a result of an MLME-BEACONNOTIFY.request. Reciept of this primitive by the ARMA serves as notification of the change of Beacon Notify as indicated.
MLME-BEACONNOTIFY.indication
This primitive reports to the ARMA that a Beacon was received on the data channel. The primitive parameters are as follows:
This primitive is generated by the MLME if a beacon is received on the data channel. Note that a separate MLME-BEACONNOTIFY.indication is generated for each beacon received, so the primitive parameter will only ever contain a single BSSDescription. Upon receipt of this primitive, The ARMA is notified of the Beacon and the signal strength at which it was received.
5.c DRCP Messages Preferred Implementation
The following describes the manner in which the above described primitives are used to implement DRCP messages in an 802.11 environment.
DRCP Preclaim
DRCP Accept
DRCP Registration Request
DRCP Registration Acknowledge
6. Movement Detection
As previously described, APs and STAs ascertain movement based upon evaluation of short and long term averages of parameters, along with expected error measurements. In accordance with an aspect of the invention, movement detection is achieved through application of a broader inventive concept that provides a way to ascertain the dynamic attributes of a system based upon short and long term averages of discrete data measurements. The principles of this invention apply to any system in which a discretely measured variable may change widely. For purposes of clarity, however, the invention is now described in terms of its particular application to wireless networks.
In a wireless network such as the one shown in
However, received power levels can appear to vary greatly even when a user is not moving. For example, the opening or closing of a doorway can cause either a gain or attenuation in the user's received power level. A person waving their hand near the user's antenna can even cause a gain or attenuation in received power level. Environmental interference can also cause changes in received power levels. These various changes in power level can cause a user to appear to be moving when in fact he or she is not. This can cause the user to roam needlessly between APs, particularly in environments where APs are close together and their transmit power levels are lower than maximum power. Alternatively, variations in signal power due to these effects can mask the fact that the user is indeed moving. In this case, the system could fail to detect the motion and fail to associate the user with the appropriate AP.
As a more particular example, consider an 802.11a wireless network. APs in such a network provide a maximum bandwidth of 54 Mbps. Bandwidth drops with distance from the AP. Assume that adjacent APs have their transmit power adjusted so that each provides a 54 Mbps cell on the order of about 10 feet in diameter. A walking user might be able to transition through such a cell in 2 seconds. On the other hand, a user sitting at his desk (near the center of the cell, right next to the AP) who gets up and leaves travels only 5 ft, not 10 ft to the edge of the cell—so, it may take the user only just over a second to be in the aisle and out of the cell. These examples provide a motivation for why rapid power estimates based on discrete measurements must be made. Increasing sampling rates increases accuracy, but this also causes more overhead in terms of wireless channel bandwidth, interrupt activity and processing overhead on the user device. Some user devices could be simple appliances such as phones, digital assistants, etc. and have very limited processing power. So a trade-off must be made between sample rate and overhead.
According to one possible implementation, power levels are measured at intervals over a window of time and a roaming decision is made. In an 802.11a network, when a user is about 1 ft from an AP whose power is set so that it has a 54 Mbps cell which is about 10 ft in diameter, the user's true mean power level should be about −38 dbm. Assume a 99% confidence interval around the true mean (i.e. the power level to be estimated) is desired. Yet, there is a variability to the measurements because of environmental effects (hand waving, etc.) as well as inherent inaccuracy in the implementation measurement itself. Assume these inaccuracies and statistical variability in the data result in distribution of the data with a standard deviation, σ=15 dbm. Referring to
The 99% confidence interval has a range of |52.6−23.4|=29.2 dbm, or about 30 dbm. This is about ±15 dbm. So, because of the variability in the signal, if only 8 samples are taken, all that can be known is that the “true power” lies somewhere between −23.4 dbm and −52.6 dbm and that such conclusion can be drawn with 99% assurance.
If less accuracy can be tolerated, for example 95% or even 90% confidence, the resultant range would be narrower. But, lower confidence intervals increase the likelihood of “false positives”. A false positive occurs when a user is ascertained to be moving when in fact he or she is not, causing the user to needlessly roam to another AP. It is desirable to minimize such false positives as they needlessly consume valuable bandwidth.
When a user doubles his or her distance from an AP, the user's received power decreases by 6 db. So, when a user moves from 1 foot away to 2 feet to 4 feet away from the AP (almost to the edge of the cell), the (true) average received power has decreased from −38 dbm to −44 dbm to −50 dbm respectively. But, as seen in
Of course increasing the number of samples taken decreases the range of error. If 20 power samples are taken, then the 99% confidence interval is −29.1 dbm to −46.9 dbm. But, taking lots and lots of samples will take too long unless channel overhead is increased.
Now consider taking n samples to produce an estimate, and then taking n more to produce a second estimate. The two estimates are then compared to see if a conclusion can be drawn as to whether the user is moving. If the confidence intervals around each estimator are large, e.g. 99%, then there exists a spectrum of outcomes and again it cannot be ascertained as to whether the user is moving toward or away from the AP, or not moving at all. In
Consider two basic scenarios regarding motion in wireless networks:
In the instance of scenario #1, which describes the very large majority of user activity in wireless networks, the accuracy of the power estimate can be greatly improved. In accordance with the principles of the invention, two averages of the received signal strength are maintained as above. But, one is the most recent N1 samples taken over a sliding window, and the other is a long term average, using N2 samples. So both a long term average and a short term average are maintained. Referring to
In
For a given application, one needs to ascertain how many samples (N2) must be taken such that the long term average estimate has essentially 0 error. Also, it is desirable to ascertain how few samples (N1) are needed in the short term average to be able to make a decision with 99% accuracy.
Assume a user starts at a position 1 foot away from the AP and moves towards the edge of the 10 ft cell. The goal is to find out how few samples are required to ascertain that the user is moving with 99% accuracy, in order to produce the most robust implementation from an overall performance perspective. If it were possible to “perfectly” measure power, a −6 db drop would be observed each time the user doubles his distance from the AP. Referring to
In this wireless network example, it has been assumed that a user can walk from the center of the cell to the edge of the cell in 1.5 seconds. So, samples need be taken every 1.5÷14≈100 milliseconds. As a further improvement, it would be desirable to use 16 samples so that division can be done by a processor via a shift operation. This increases computational efficiency on the user's machine. This increases the sample rate a negligible amount.
Regarding the long term average, it may be reasonable to tolerate a ±1 db confidence interval around the long term estimate. The tighter this interval needs to be, the longer the user has to stay near the AP, or stay relatively stationary within a certain area, to cause the average to converge. It is desirable to calculate how little time the user needs to stay in place to achieve an accuracy of ±1 db with 99% confidence. Assume, reasonably, that signal strength samples are taken based on messages received from the AP every 50 milliseconds. Referring to the table shown in
The preferred implementation for the current wireless network example thus utilizes: (a) a short term average over N1=16 samples, (b) a long term average over 2048 samples, and (c) “N2” which is the number of samples taken so far in computing the long term average. The process is as follows:
In an environment where users tend to remain in a cell for less than 1 second, the long term estimate could be used based on fewer samples. However, this will result in an increased risk of false positives. Several alternatives can be considered to mitigate the occurrence of false positives:
To generalize, in a system wherein particular dynamic attributes are to be ascertained (e.g. “is the wireless network user in motion”), short term and long term averages of a system variable (e.g. signal strength) are calculated. An acceptable difference between the short and long term averages is calculated which positively identifies the system characteristic (e.g. the user has moved.)
Referring to
In
In
In
A short term average 656 and long term average 658 are calculated based on the contents of the respective ring buffers 652 and 654. A comparator 660 uses a stored allowed range 662 and the short term average 556 and long term average 658 to produce the movement indication 650 in accordance with the process of
In
Though the above described aspects of the invention have been exemplified as they apply to wireless networks and, in some particularity, 802.11 networks, it will be clear to the skilled practitioner that the invention can be employed in any wireless communications environment, including wireless data networks, wireless phone networks, and wireless I/O channels. All aspects of the invention may be implemented in either hardware or software. The preferred embodiment has been described as a software architecture because of its advantageous ease of portability between various hardware platforms.
It will be understood that each block of the flowchart illustrations, and combinations of blocks in the flowchart illustrations, can be implemented by computer program instructions. These computer program instructions may be loaded onto a computer, embedded device microprocessor (such as that found in an AP or STA), or other programmable data processing apparatus to produce a machine, such that the instructions which execute on the computer or other programmable data processing apparatus create means for implementing the functions specified in the flowchart block or blocks. These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart block or blocks. The computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks.
Those skilled in the art should readily appreciate that programs defining the functions of the present invention can be delivered to a computer in many forms; including, but not limited to: (a) information permanently stored on non-writable storage media (e.g. read only memory devices within a computer such as ROM or CD-ROM disks readable by a computer I/O attachment); (b) information alterably stored on writable storage media (e.g. floppy disks and hard drives); or (c) information conveyed to a computer through communication media for example using baseband signaling or broadband signaling techniques, including carrier wave signaling techniques, such as over computer or telephone networks via a modem.
While the invention is described through the above exemplary embodiments, it will be understood by those of ordinary skill in the art that modification to and variation of the illustrated embodiments may be made without departing from the inventive concepts herein disclosed. Moreover, while the preferred embodiments are described in connection with various illustrative program command structures, one skilled in the art will recognize that the system may be embodied using a variety of specific command structures. Accordingly, the invention should not be viewed as limited except by the scope and spirit of the appended claims.
| Number | Date | Country | |
|---|---|---|---|
| 60449602 | Feb 2003 | US | |
| 60466448 | Apr 2003 | US | |
| 60472320 | May 2003 | US | |
| 60472239 | May 2003 | US |
