Patent Application
20050180356
1. Field of the Invention
The present invention relates generally to wireless networks and, more particularly, to a protocol for facilitating communication in a self-organizing, wireless network.
2. Background
The information contained in this section relates to the background of the art of the present-invention without any admission as to whether or not it legally constitutes prior art.
Wireless networks that are often referred to as ad hoc networks typically lack a fixed, pre-existing infrastructure such as base stations and access points. Thus, these networks typically lack the rigid master (clocker) and slave (clock follower) relationships present in other networks. A network wherein any node can provide the communications clock is referred to as a “peer-to-peer” network.
Further, many of these ad hoc networks operate in the unlicensed radio frequency bands, resulting in the potential of interference with other networks.
Following the IEEE 802 network convention, the Data Link Layer of the OSI Reference Model is divided into two sublayers: the Logical Link Control (LLC) layer and the Media Access Control (MAC) layer. The MAC layer interfaces directly with the network media.
Many MACs that appear in radio systems lack important features to support direct peer-to-peer communication. One problem with broadcasting on a particular channel is that every node in the network must listen to the same channel at the same time.
On the other hand, it may be desirable to hop among many channels and avoid interference with a peer-to-peer MAC. Peer-to-peer MACs that require exhaustive scanning of a large set of channels typically have low data throughput.
Peer-to-peer MACs that rely upon a single channel are unreliable when there is channel contention (such as that from microwave ovens, or from IEEE 802.11b wireless systems or cordless phones.)
No practical solutions to the problems of a multi-channel broadcast peer-to-peer MAC have appeared in the literature. As a result, existing MACs typically use a single channel or a single hopping sequence in the case of a frequency-hopping network, such as a network operating under the Bluetooth protocol. For further information on the Bluetooth protocol, reference may be made to http://www.bluetooth.com/pdf/Bluetooth—11_Specifications_Book.pdf.
Another problem in broadcasting is that known as the “hidden node” problem. The “hidden node” problem refers to a scenario in which packet collisions may occur when two transmitting nodes are unaware of each other. For example, node X may want to send a packet to node Y, and node Z may also want to send a packet to node Y. Node Z and node X may be too far apart to be able to detect each other. This may cause a packet collision at node Y. If this is a problem for unicast messages, it is even more of a problem for broadcast messages, because broadcast packets are often used to “flood” a network.
No practical solutions to the hidden node problem for wireless broadcasts have appeared in the literature. As a result, broadcast transmissions are always less reliable than unicast transmissions in wireless networks. In a CSMA/CD Ethernet network, broadcast transmissions are exactly as reliable as unicast transmissions. Therefore, any wireless self-organizing routing system that relies on broadcast capability to maintain network connectivity and routing information is inherently less reliable than a wired routing system such as Ethernet
The disclosed embodiments described herein overcome the foregoing problems. The disclosed embodiments of the present invention include a multi-hop routing architecture to configure an ad hoc network automatically and to transport data. A node in the network is capable of sending and receiving control data and payload. A self-organizing capability allows a network to adapt to the changing network topology and RF reception conditions, for example, and to reconfigure itself dynamically to route packets around troubled areas. It also allows a network to expand an existing installation easily and reliably without relocating any existing “hub” nodes. In addition, disclosed embodiments of a multi-hop network save power and extends the reach of a network, without adding expensive wires or amplifiers.
The disclosed embodiments of the multi-channel broadcast MAC described herein have a reliable broadcast mechanism that solves the hidden node problem and channel-hops when necessary. A multi-hop, ad hoc routing network can thus be constructed reliably and with greater bandwidth efficiency using this broadcast paradigm. Specifically, the disclosed embodiments of the multi-channel broadcast MAC resolves the problems encountered in wireless communications, including hidden and exposed node, collision during mastering of a channel, channel contention bottlenecks in peer-to-peer communications, interference or jamming on a broadcast channel, and low-duty cycle communications to prolong battery life.
The disclosed embodiments include a multi-channel peer-to-peer broadcast MAC that supports multiple channel masters. The disclosed embodiments of the MAC have a reliable broadcast that solves the hidden node problem and channel-hop around interference.
When transporting unicast messages, certain disclosed embodiments of the MAC perform “Area Load Balancing” (ALB) to balance wireless throughput in a geographic area, for example. The disclosed MACs may carry packets of arbitrary size. Disclosed embodiments of a multi-hop routing engine built upon the disclosed broadcast MACs may allow a plurality of network nodes to discover all their neighboring nodes automatically and to organize themselves reliably into a communications network. When a network element is relocated or when one or more connections or paths fail, the disclosed routing engine can rapidly reorganize the network topology and restore the network connectivity between communicating peers.
To minimize mutual interference in the shared-use bands, ad hoc networks must generally perform adaptive channel selection. One aspect of this invention allows dynamic channel selection even during the transmission of broadcast packets.
This summary does not purport to define the invention. The invention is defined by the claims.
In the following description, for purposes of explanation and not limitation, specific details are set forth in order to provide a thorough understanding of the present invention. However, it will be apparent to one skilled in the art that the present invention may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known methods and devices are omitted so as to not obscure the description of the present invention with unnecessary detail.
The Network layer 16 performs multi-hop routing to provide the functional and procedural means of transferring variable length data sequences, for example, from a source to a destination. The Network layer 16 is provided with a routing table 28 for facilitating the routing of packets. The structure of the routing table 28 is described in detail below with reference to
The Routing module 18 may establish and maintain the topology and routing information. In this regard, the routing module 18 may be responsible for updating the topology and routing information on, for example, a predetermined periodic or an as needed basis.
The Control and Management module 20 may provide error reporting for remote maintenance, diagnosis, and administration. The Application protocol module may be a user or machine process that makes use of the reliable multi-hop routing engine to provide communications among the nodes in a wireless network.
The Media Access Control (MAC) sublayer 12 includes two communication modules, a broadcast module 24 and a unicast module 26, the functions of each of which are described below in further detail.
The Internet IP packet 30 is encapsulated in a Network Layer Body 36 of the Network layer 16. The Network Layer Body 36 contains the protocol transport information. In addition to the Network Layer Body 36, the Network layer 16 includes a Protocol block 38. The entry in the Protocol block 38 is one of the pre-assigned protocol numbers. For example, 0×01 is for Routing protocol, 0×03 is for the 802 envelope, 0×04-0x0F are reserved for system protocols, and 0x10-0xFF are for application protocols.
The Network layer 16 also includes a Packet ID 40, which may be a natural integer, to identify a packet. The Packet ID 40, together with an Original Source Address 48 included in the Network layer 16, produces a unique transmission ID that is used to suppress duplicate broadcasts. Applications may also use this field to suppress duplicate transmissions.
A Time-To-Live (TTL) counter 42 and an Original TTL 44 are also included in the Network layer 16. The TTL counter 42 starts at some well-known value that may be set according to the size of the network, for example, and is decremented every time a node forwards the packet to another node. After repeated forwards, the TTL counter 42 may reach zero before reaching its intended destination. In this case, the Network layer 16 discards the frame, and the Control and Management 20 (shown in
The Original TTL 44 is the Original Time-To-Live counter. The Original TTL 44 describes the initial value of the TTL 42 field. The Original TTL 44 and the TTL 42 are noted by the Routing module 18 (shown in
The Original Destination Address 46 and the Original Source Address 48 are supplied by the client and represent the end-to-end addresses. In one embodiment, these addresses follow IEEE addressing conventions, such as Organizationally Unique Identifiers (OUI).
The Network layer 16 is embedded in a MAC body portion 50 of the MAC layer 12. In addition to the MAC body portion 50, the MAC layer 12 also includes a destination address 52 and a source address 54 for the present hop. These addresses 52, 54 are not necessarily the same as the Original Destination 46 and Original Source 48 addresses in the Network Layer. In fact, except for a single hop transmission, the two sets should never be identical.
The source address 54 is the sender's address for this hop (local) and follows IEEE conventions. The source address 54 should not be a broadcast address because of the danger of creating a broadcast storm.
The MAC layer 12 also includes a Type field 56.
The MAC body portion 50 is followed by a CRC field 58. The CRC field 58 contains a cyclic redundancy check code for data error detection.
In a multi-hop network, the original client may simply fill in an Original Destination address 44 and Original Source address 48, and the Type field 56 for the packet. The Network layer 16 will fill in the rest of the packet. In this arrangement, the addresses are typically 48-bit IEEE 802 compatible MAC identifiers, or OUI's.
In normal operation, the various nodes of a network may follow a priority-driven algorithm that consists of three steps. The first step may be to scan a set of channels to detect potential transmissions. The second step (if the scan detected a message) is to receive a message. The third step is to attempt to transmit a message if one exists. Whenever a node finishes transmitting or receiving it always returns immediately to the channel scanning state.
Potential transmitters play a repeating “Request to Send” (RTS) message header on a channel to indicate a desire to transmit. Scanning nodes first poll a series of m fixed, well-known channel numbers which are preferably spread evenly throughout a communications band. Then the scanning nodes poll a second series of n channels that are also spread throughout the communications band but they are not well-known and depend upon a function of the scanning node's unique identification number (e.g. an OUI or 48-bit IEEE 802 MAC ID.)
In this arrangement, m fixed broadcast channels are used to allow frequency diversity. To implement broadcast, a node X first determines that no node has a broadcast channel, and then it masters a broadcast channel and places a repeating message on the channel. This message consists of the node's unique identification and an RTS message header. As slaves lock onto the master node X, if a slave node Y sees the RTS, it turns on its transmitter and sends a Clear to Send (CTS) message, completing a CSMA/CA transaction. When the master sees the CTS message it converts its RTS message into a RTS Extension (RTSX) message. This message has not appeared in previous systems. When this conversion occurs, the transmitting slave becomes the “primary slave.” Other slaves that arrive later are referred to as “secondary slaves.” If this confirmation does not come within a given timeframe (e.g., 10 ms), the broadcaster node X assumes that there is either a master collision, slave collision, or interference. Subsequently, node X stops the current attempt and hops to another broadcast channel to begin another attempt.
In one embodiment, the communications channel is TDMA multiplexed, thereby both the master and slave nodes can transmit in alternating time slots. An RTS or CTS message fits in exactly one slot, although this is not a requirement.
Next, the polling algorithm may set out to select a particular channel for communication. A node wishing to communicate may poll the first broadcast channel, CH-160, to determine the relative signal strength, for example, by sensing the pre-carrier on the channel (block 72). This is generally performed by the receiver component of the node. If the signal strength is unsatisfactory, broadcast channel CH-I is skipped (line 74), and scanning is advanced to broadcast channel CH-262.
If the signal strength at block 72 is satisfactory, as indicated by line 76, the process continues to determine if a lock can be achieved by sensing the carrier (block 78) on channel CH-160. If a lock cannot be achieved, channel CH-1 is skipped (line 80), and scanning is advanced to channel CH-262.
If a lock is achieved at block 78, as indicated by lines 82 and 84, the process senses whether there are any unicast RTS messages (block 86) or multicast RTS messages (block 88) on channel CH-160. If neither exists, channel CH-160 is determined to be free, and the node can proceed with its broadcast channel setup. Otherwise, the scanning proceeds to channel CH-262.
If for any reason, channel CH-160 is determined to be unavailable, the entire process is repeated at channel CH-2. The process continues to each successive lower prioritized channel until a satisfactory channel is found. If all M broadcast channels have been exhausted without finding a satisfactory channel, the broadcast packet is discarded.
A node X wishing to send a unicast message to node Y calculates a hash function, HASH(Y, I), using Y's OUI and 1=0, 1, . . . , n−1, for example, to determine which of the several (e.g., n) unicast channels to use. Node X then performs channel-sensing on channel HASH(Y, 0). If the channel is clear, node X sends out the desired “Request To Send” signal, (RTS, X), and waits for an acknowledgement by way of a “Clear to Receive” signal (CTS, X). If there is no CTS within a timeout period, a reception SCAN is performed. The reception scan includes scanning the channels for possible signal reception. If the reception SCAN fails, then node X performs channel sensing on channel HASH(Y,I), and so forth. In this way, peer-to-peer communications can occur in parallel in the network.
As an example, a chip, such as a cordless phone chip, can support 54 channels with n equal to four. In this example, a number between 0 and 12 is generated by hashing the OUI of node Y, and this is added to CHANNEL(I)=0, 14, 27, or 41 to obtain a destination channel HASH(Y, I)=HASH(Y)+CHANNEL(I). For the initial attempt, node X masters channel HASH(Y, 0) sending out the desired (RTS, X), and waits for acknowledgement (CTS, X). If there is no CTS after a timeout period, the transmitter tries other channels until a (CTS, X) is received by the master node. Once the CTS is received, the master node can then proceed with the unicast operation by sending one or a sequence of data packets.
The multi-channel broadcast MAC provides load balancing for unicast messages in the network. For example, whenever node A wants to transmit to node B, and node C wants to transmit to node D, a hash function is applied to the destination addresses to determine a set of channels to use. There will be no interference if Hash(B,i)!=Hash(D,j). In one embodiment, there are 13 hash equivalency classes. Thus, a conflict only occurs if i=j and HASH(B)=HASH(D). This technique makes it possible for a peer-to-peer network to, for example, carry up to 13 times more data because different pairs of nodes will use separate channels for unicast communications and results in a reduced probability of interference.
As shown in
Once an RTS/CTS handshake (96, 98) has successfully occurred, a unicast data transmission sender can proceed with the operation by sending one or more DATA packets 100, 102, 104 on a unicast channel. For broadcast operation, successful reception of a CTS 98 causes the broadcaster node X 92 to change its message on the broadcast channel to an RTSX message 106.
For broadcast operation, after sending the RTSX message 106, the broadcaster waits for a certain amount of time for other channel-hopping nodes to find the RTSX channel and lock on. Once the RTSX message 106 occurs, additional nodes may detect the RTSX message 106 and settle on the broadcast channel. In certain embodiments, these secondary slave nodes may also turn on their transmitter to produce a spatial reservation for their geographic area. In this regard, all slave nodes may radiate RF power in the same band or time slot using the same PN codes to accomplish this reservation. Because all slaves turn on their transmitters, hidden nodes are able to detect and avoid the reserved broadcast channel.
Finally, after a period of time, the broadcaster node X sends a data packet or a sequence of data packets 100, 102, 104. The time period may be predetermined and may be based on the relative distance between the nodes or the size of the network. The primary slave node Y 94 receives the data packets and verifies a checksum and returns either an acknowledgement (ACK) or a negative-acknowledgement (NACK) message 108 to indicate where or not the correct message was received. At the same time the secondary slaves turn off their transmitters so that the primary slave may send an acknowledgement (ACK) message (not shown).
In one embodiment, the RTSX message 106 may contain a repeating countdown timer that indicates how long to wait before the data packets 100, 102, 104 begin to arrive. In this embodiment, the secondary slaves may turn off their transmitters at the right time, even if they are temporarily unable to hear the transition from the RTSX message 106 to data packets 100, 102, 104. This prevents the secondary slaves from inadvertently jamming the acknowledgement from the primary slave node Y.
In another embodiment, the secondary slave nodes may not turn on their transmitters in order to avoid disturbing the locking algorithm between the primary slave node Y 94 and the master node X 92.
To solve the hidden node problem for broadcast in a system with multiple channels, an RTS/CTS transaction using the carrier alone can be performed. As a requirement, all nodes in the network must have an ability to transmit a constant carrier. In this arrangement, a new node that has just hopped onto a channel that has the RTS or CTS carrier signals can immediately defer until the communications is finished. This mechanism is effective even in scenarios in which newly arriving nodes can only sense the presence of the slaves such as node Y.
To initiates a transmission operation, node X 110 may first determine that no node has a broadcast channel. This determination may include failure to detect a broadcast. Node X 110 then masters a broadcast channel and places a repeating RTS message 126 that repeats for a predetermined number of periods. When node Y 112, being in node X's range, detects the RTS message 126, it turns on its transmitter and sends a CTS message 128 for a predetermined number of periods to complete a handshake. Once the handshake has occurred, unicast data transmission can proceed immediately. For broadcast mode, node X 110 changes the RTS message 126 on the broadcast channel to an RTSX message 130 and repeats it for a predetermined number of periods. As described above with respect to
When node A 114 finds the RTSX channel and locks onto node X 110, node X 110, node Y 112, and node A 114 are all generating RF signals in their respective geographic regions. The presence of RTS, CTS, or RTSX messages causes node A 114 and other nodes to avoid transmission. Because these carrier signals are generated quasi-continuously, if node Z 116, for example, is channel hopping, it will also avoid transmitting data when it arrives at the channel used by nodes X 110 and Y 112. This therefore solves the hidden node problem for broadcast.
In one embodiment, two forms of link sensing may be used to implement the solution to the hidden node problem: Relative Signal Strength Indicator (the electromagnetic field strength detected by the radio) and Link Quality (a number generated by the radio that describes how many bits are corrected within 64 bit payload, for example).
By solving the hidden node problem, broadcast transmissions may give the same guarantees as unicast transmissions, that at least one node received the packet successfully. This may be important in a network with dynamic routing, because broadcast messages are the backbone of any self-organizing routing system.
One critical aspect of an ad hoc network as described above is the network's ability to self-organize. In this regard
For a new, non-duplicate broadcast, the node 138 recognizes the “Original Source Address” in the packet as the address of the root node 136. On the other hand, the “Source Address” indicates the immediately previous node in the path, node 140, and, therefore, the preferred route to the root node 136. The Routing module at the node 138 stores the “Original Source Address” of the packet into the “Destination Address” field of a routing table entry, an example of which is described below with reference to
After updating the routing table entry for the root node 136, each node, including node 138, replaces the “Source Address” entry in the packet with its own address and rebroadcasts the packet. The solid arrows in
The root node 144 may then update the entry for the remote node 146 in a routing table by storing the entry in the “Original Source Address” field of the packet into the “Destination Address” field, and the entry in the “Source Address” field into the “Next Hop Address” field of the routing table entry. A next “local hop” route from the root node 144 to the remote node 146 is thereby created or updated in the routing table. Similar routing table entries may be created or updated for each remote node in the network 142. Solid arrows in
The multi-hop routing engine uses the foregoing MAC paradigm for enhanced reliability. In this arrangement, routing of data packets is performed on a hop-by-hop basis. Specifically, a routing table is maintained at each node in a network. Two aspects concerning network topology discovery include Root Synchronization and Route Learning.
Root Synchronization, described above with reference to
Route Learning, described above with reference to
The Routing layer may not only recognize broadcasts and forward them, it may also perform duplicate suppression so that the network is not flooded with a large number of broadcast packets. Duplicate suppression can be performed using a table of recently forwarded broadcast packets and the best TTL learned to date.
The protocol described above allows a tree-based routing topology to be constructed in response to a single, root-synchronization packet. This tree-based topology minimizes memory consumption in the intermediate nodes. In many applications each node of the wireless network may send and receive packets to the Internet and not to adjacent or non-adjacent nodes. For these applications, a tree-based routing protocol may be necessary and sufficient. On the other hand, if a workstation user roams through the network and wishes to communicate with every node, he may also send root synchronization packets and collect the network topology in a series of response packets. In a third example, a node may need to communicate only with its neighbors which are, for example, two or three hops away. In this case, the node may send root synchronization packets with a limited time to live. Thus, the disclosed protocol provides a tremendous degree of flexibility in a network where each node has a limited memory space for routing tables.
Once the routing table is constructed, it may be used each time a packet is forwarded to determine its next hop or whether it has reached its final destination. Specifically, a node may first determine if the packet has reached its final destination or the TTL value has become zero. If the packet has not reached its final destination and the TTL value is greater than zero, the node may use the entry in the “Original Destination” field of the packet header as a key to search the routing table. If a matching table entry is found, a “Next Hop” field of the matched table entry may be used to modify the packet header. For example, the entry in the “Source” field may be changed to the current node's address, and the TTL field may be decremented by one. The packet may then be transmitted to its next hop en route to the Original Destination node.
As described above, each node in the network preferably maintains a routing table.
It will be apparent to those skilled in the art that the disclosed embodiments of the MAC and multi-hop routing networks provide a number of important advantages. At the media access layer, the multi-channel broadcast MAC provides a reliable, load-balanced and high-throughput broadcast. This enhances the reliability of a multi-hop routing network and allows the network to adapt to the changing network topology and RF-reception conditions. At the routing layer, only nodes that need full topology information need to broadcast the root synchronization packets. This keeps routing table sizes to a minimum. Further benefits of the disclosed embodiments of a dynamic routing protocol allow a rapid restoration of connectivity in case of node or path failure.
It will be evident that the benefits of this invention also apply to other type of networks.
While particular embodiments of the present invention have been disclosed, it is to be understood that various different modifications and combinations are possible and are contemplated within the true spirit and scope of the appended claims. There is no intention, therefore, of limitations to the exact abstract or disclosure herein presented.
| Number | Date | Country | |
|---|---|---|---|
| 60415056 | Oct 2002 | US |
