This invention is being co-filed on the same day as another application titled “Deployment Strategy For Sensors With Sensing And Sensed Regions” by present inventors Hector H. Gonzalez-Banos and Asif Ghias.
This invention relates generally to the fields of computational geometry, combinatorics, set theory, linear programming, computer science, distributed/mobile sensor networks, wireless sensor networks, smart sensor networks and in particular to determining effective placement of different types of sensors in varied environments.
There are a number of related disciplines with similar and sometimes conflated names such as wireless sensor networks, distributed sensor networks, mobile sensor networks, ubiquitous sensor networks, smart sensor networks that are concerned with effectively deploying various types of sensors in diverse environments for a large variety of industrial applications. It is no surprise that sensor deployment in such disciplines remains an active area of academic and industrial pursuit. With the ubiquity of sensors such as smart phones and other smart devices pervading through our daily lives, with concepts such as internet of things (IOT) maturing over the last decade, and with the interconnectedness of the world fast becoming a reality, it is no surprise that a large number of technology companies and academic institutions are spending a vast amount of resources in developing programs and products for deploying the ever increasing universe of sensors in the most effective manner possible.
In as far as devising strategies for deploying sensors, there are many schemes taught in the prior art. “A Randomized Art-Gallery Algorithm for Sensor Placement” by Hector Gonzalez-Banos et al. of Stanford University (2001) describes a placement strategy for computing a set of ‘good’ locations where visual sensing will be most effective. The sensor placement strategy relies on a randomized algorithm that solves a variant of the art-gallery problem known to those skilled in the art. The strategy finds a minimum set of guards inside a polygonal workspace from which the entire workspace boundary is visible. To better take into account the limitations of physical sensors, the algorithm computes a set of guards that satisfies incidence and range constraints.
“Coverage by directional sensors in randomly deployed wireless sensor networks” by Jing Ai et al. of Rensselaer Polytechnic Institute (2005) teaches a novel ‘coverage by directional sensor’ problem with tunable orientations on a set of discrete targets. It proposes a Maximum Coverage with Minimum Sensors (MCMS) problem in which coverage in terms of the number of targets to be covered is maximized whereas the number of sensors to be activated is minimized. The paper presents its exact Integer Linear Programming (ILP) formulation and an approximate (but computationally efficient) centralized greedy algorithm (CGA) solution. These centralized solutions are used as baselines for comparison. Then it provides a distributed greedy algorithm (DGA) solution. By incorporating a measure of the sensors residual energy into DGA, it further develops a Sensing Neighborhood Cooperative Sleeping (SNCS) protocol which performs adaptive scheduling on a larger time scale. Finally, it evaluates the properties of the proposed solutions and protocols in terms of providing coverage and maximizing network lifetime through extensive simulations.
“Selection and Orientation of Directional Sensors for Coverage Maximization” by Giordano Fusco et al. of Stony Brook University (2009) addresses the problem of selection and orientation of directional sensors with the objective of maximizing coverage area. Sensor nodes may be equipped with a ‘directional’ sensing device (such as a camera) which senses a physical phenomenon in a certain direction depending on the chosen orientation. The paper addresses the problem of selecting a minimum number of sensors and assigning orientations such that the given area (or set of target points) is k-covered (i.e., each point is covered k times). The above problem is NP-complete, and even NP-hard to approximate. The paper presents a simple greedy algorithm that delivers a solution that k-covers at least half of the target points using at most M log(k|C|) sensors, where |C| is the maximum number of target points covered by a sensor and M is the minimum number of sensors required to k-cover all the given points.
In “Efficient Sensor Placement for Surveillance Problems”, Agarwal et al. of Duke University (2009) studies the problem of covering by sensors of a two-dimensional spatial region P that is cluttered with occluders. A sensor placed at a location p covers a point x in P if x lies within sensing radius r from p and x is visible from p, i.e., the segment px does not intersect any occluder. The goal is to compute a placement of the minimum number of sensors that cover P. It proposes a landmark-based approach for covering P.
In “On Sensor Placement for Directional Wireless Sensor Networks”, Osais of Carleton University, Ottawa (2009) discusses a directional sensor network that is formed by directional sensors which may be oriented toward different directions. The sensing region of a directional sensor can be viewed as a sector in a two-dimensional plane. Therefore, a directional sensor can only choose one sector (or direction) at any time instant. They discuss the placement of such directional sensors as a critical task in the planning of directional sensor networks. They also present an integer linear programming model whose goal is to minimize the number of directional sensors that need to be deployed to monitor a set of discrete targets in a sensor field. Numerical results demonstrate the viability and effectiveness of the model.
In general, the problem of sensor placement in an occluded workspace is well studied. Such a system 10 of prior art is illustrated in
Further, a system 20 of prior art is illustrated in
A shortcoming of prior art teachings is that they do not provide a strategy for sensor deployment that includes multiple sensors or sensing stations, each with different sensing/visibility models and constraints. Further, the prior art assumes a simple sensing model for the sensors that is a based on an individual type of sensor, rather than a composite visibility model that is based on a collection of various types of sensors on a given sensing station. A further shortcoming of the prior art is that it generally conflates the notions of ‘sensing coverage’ that is concerned with sensing a set of target sites or other sensors or sensed stations in a workspace, and ‘network coverage’ that is concerned with connecting or communicating with the target sites or other sensors or sensed stations in the workspace, sometimes using a different type of sensor or transceiver.
The prior art teachings are also silent on the notions of ‘localizability coverage’, which we refer to as the capability to determine the location of a sensed station by knowing the positions of two or more sensing stations—as will be taught in the detailed description section. Further, the prior art does not treat the notion of a sensed station with its own composite sensed region as a collection of individual sensed regions, rather than just a site or location of interest in the workspace.
In view of the shortcomings of the prior art, it is an object of the present invention to teach a more effective deployment strategy for sensors than is available through the teachings of the prior art.
It is further an object of the invention to allow sensing stations having multiple types of sensors, each with its own sensing model and constraints.
It is further an object of the invention to incorporate sensing coverage, network coverage and localizability coverage simultaneously in the deployment of sensors as taught by the present invention.
It is further an object of the invention to allow sensed stations having multiple types of sensors, each with its own sensed model and constraints.
The objects and advantages of the invention are given by a system and methods for determining a set of placement sites from a set of candidate sites in a workspace. The candidate sites refer to the potential locations and other configuration information of sensing stations in the workspace, while the placement sites determined or computed by the system comprise those candidate sites where the sensing stations should be placed or deployed in order to ensure coverage. The system further comprises a set of target sites in the workspace. The target sites refer to the potential locations and configurations of sensed stations in the workspace. The system further comprises zero or more obstructions in the workspace that would obstruct the sensing of the sensed stations by the sensing stations in the workspace.
Each sensing station has one or more sensing regions around it, each such sensing region likely but not necessarily existing due to individual sensors on the sensing station. The sensing region is defined as the set of all sites within the workspace that are able to be sensed by that sensing station in the workspace, despite the obstructions. The sensing region is further constrained by a sensing range and a sensing orientation of the corresponding sensing station. Subsequently, a composite sensing region for each sensing station is defined as the collection of the individual sensing regions of the sensing station.
Corresponding to a set of candidate sites, a set of ranges having the same size as the set of candidate sites is defined. Each range in the set of ranges is selected to be the subset of the set of target sites such that the sensed stations at the target sites in the subset are all able to be sensed by the sensing station at that candidate site corresponding to the range. As stated, the cardinality of the set of ranges is the same as the set of candidate sites being considered. The system determines the near-optimal set of such candidate sites in the workspace to be the placement sites (or the computed solution) based on a minimum set-cover solution for the set system comprising the set of all target sites, and the set family comprising the set of all the ranges above.
In the preferred embodiment, the collection of individual sensing regions of a sensing station is taken to be a union of the individual sensing regions. Alternatively, the collection is taken to be an intersection of the individual sensing regions. Still in another embodiment, the collection is taken to be based on a generic set operation of the individual sensing regions of the sensing station.
Preferably, the target sites are merely the locations of interest that need to be observed in the workspace. Hence, there is no sensor or device present at the location that needs to be observed in the workspace in such a preferred embodiment. In another preferred embodiment, the set of placement sites determined by the system guarantees that each sensed station is able to be sensed by at least two sensing stations at a given point in time. This capability affords the determination of the location of a sensed station or a target site by triangulation. If each sensed station is able to be sensed by three or more sensing stations then this capability affords the determination of the location of a sensed station or target site by trilateration.
In another preferred embodiment, the candidate site comprises the location of the site in two or three dimensions in the workspace, and the angle(s) of orientation of the sensing station placed at that candidate site in two or three dimensional Euclidean space respectively. Preferably, the angle(s) of orientation is/are unconstrained or omni-directional, so that the candidate site merely refers to the location of the placement of the sensing station.
Similarly, in another preferred embodiment, the target site comprises the location of the site in two or three dimensions in the workspace, and the angle(s) of orientation of the sensed station placed at that candidate site in two or three dimensions respectively. Preferably, the angle(s) of orientation is/are unconstrained or omni-directional, so that the target site merely refers to the location of the placement of the sensed station.
In a highly preferred embodiment, each sensed station also has one or more sensed regions around it, likely but not necessarily, as a result of the individual sensors present on the sensed station. A sensed region of a sensed station at a target site is defined as the set of all sites around that sensed station that if overlapping with the sensing region of a sensing station at a candidate site in the workspace will result in that sensed station being sensed by that sensing station. Preferably, the sensing region of a sensing station at a candidate site and a sensed region of a sensed station at a target site are defined such that the sensing station can sense the sensed station only if the sensed station is in the sensing region of the sensing station and the sensing station is in the sensed region of the sensed station.
Preferably, the sensed region is further constrained by a sensed range and a sensed orientation of the sensed station. Preferably, there is a composite sensed region around each sensed station that is a collection of the individual sensed regions around the sensed station. Preferably, each range in the set of ranges above is further chosen such that the candidate site corresponding to the range lies in the sensed region of the sensed stations placed at the target sites of that range.
In the preferred embodiment, the collection of individual sensed regions of a sensed station is taken to be a union of the individual sensed regions. Alternatively, the collection is taken to be an intersection of the individual sensed regions. Still in another embodiment, the collection is taken to be based on a generic set operation of the individual sensed regions of the sensed station.
Preferably, the candidate sites determined by the apparatus and methods of the invention overlap with the target sites. Alternatively, the candidate sites determined by the invention do not overlap with the target sites. In another advantageous embodiment there is another constraint placed on the apparatus and methods of the invention that requires the placement sites to be have locations chosen from a set of predetermined locations in the workspace. Similarly, in another embodiment the constraint placed is such that the placement sites can only be chosen from a predetermined region in the workspace. Still in another preferred embodiment, there is a predetermined number of a certain type of sensors available.
In a highly preferred embodiment, the set-cover solution determined by the invention is based on the popular Greedy algorithm. Preferably the minimum set-cover solution is derived in polynomial time. Still preferably, the solution derived is of the order in Big-O notation at most a factor of O(d log dC*) from its optimal size C*, where d denotes the Vapnik-Chervonenkis dimension (VC-dimension) of the set system comprising the set of all target sites and the set family comprising the set of ranges. Still preferably, the VC-dimension is bounded by O(log h) where h represents the number of obstructions in the workspace.
In a highly preferred embodiment, the sensing station is a camera and the target sites comprise a surveillance space that needs to be monitored. In another preferred embodiment, the sensing stations and the sensed stations are wireless sensors operating substantially in the popular 60 GHz frequency range. In still another preferred embodiment, a sensing station is a three-dimensional object in a video, whether the video is pre-recorded or streaming, while the workspace itself comprises the video. In another preferred embodiment, the sensing stations and sensed stations are people, the workspace is a geographical place or terrain, and the candidate and target sites are the coordinates of the locations of the particular sites where the above people i.e. sensing and stations respectively, are to be located in the given workspace or the geographical place or the terrain.
In a highly preferred set of embodiments, a sensing station is a person while the workspace comprises a social graph. In a variation of the same embodiment, the sensing station is a product while the workspace comprises a social graph.
The figures and the following description relate to preferred embodiments of the present invention by way of illustration only. It should be noted that from the following discussion, alternative embodiments of the structures and methods disclosed herein will be readily recognized as viable alternatives that may be employed without departing from the principles of the claimed invention.
Reference will now be made in detail to several embodiments of the present invention(s), examples of which are illustrated in the accompanying figures. It is noted that wherever practicable, similar or like reference numbers may be used in the figures and may indicate similar or like functionality. The figures depict embodiments of the present invention for purposes of illustration only. One skilled in the art will readily recognize from the following description that alternative embodiments of the structures and methods illustrated herein may be employed without departing from the principles of the invention described herein.
The present invention will be best understood by first reviewing the sensor deployment system 100 illustrated in
Workspace 102 has six obstructions 104 as indicated in
Throughout the following explanation workspace 102 will be assumed to comprise of a collection of sites embodied by its interior, excluding obstructions 104. Generally, a site will represent a physical location in the workspace and additional configuration information of the sensor present at that location. The terms, a site, a point or location may be used interchangeably in the following explanation, and distinction between them will be drawn where needed and appropriate. Furthermore in the following text, where appropriate, the term sensor may be used to refer to a sensing station as well as a sensed station, when the distinction between the two is obvious from the context.
Sensor deployment system 100 of
According to the invention, a set of target sites in workspace 102 represents the locations of interest that are required to be observed. There are sensed stations (not shown) placed at the target sites that are sensed by sensing stations. Preferably, the sensed stations merely represent the sites or locations in workspace 102 that are required to be observed. Such a preferred embodiment is illustrated in
According to the apparatus and methods of the main embodiments of the present invention, a sensing region exists around each sensing station when that sensing station is at a given site, called candidate site in workspace 102. A candidate site would generally comprise the location information of the sensing station in workspace 102, the type of sensing station (106, 108 or 110) and any other ancillary information that may be needed to be associated with the candidate site. Such ancillary information may include, but is not limited to, the configuration of the sensor including its orientation, its sensing region (as will be taught below), and its any other capabilities or constraints, etc. Note, the invention refers to all the such potential sites where a sensing station can be placed in the workspace as candidate sites, and it refers to that subset of candidate sites where the sensing stations should be placed or deployed in order to ensure coverage, as placement sites. In other words, the set of placement sites or simply placement sites refers to the ‘computed solution’ of the sensor deployment strategy as offered by the instant invention.
The location information of a candidate site may include the two-dimensional or three-dimensional coordinates in a two-dimensional or three-dimensional Euclidean space of the system. The orientation information of a sensing station may include its three axes of orientation with respect to a given coordinate system. The orientation may be represented by rotation matrices Rx(∝) for rotation by angle ∝ around x-axis, Ry(β) for rotation by angle β around y-axis and Rz(γ) for rotation by angle γ around z-axis, or by Euler angles or still by any other rotation convention familiar to people of skill.
Further, a skilled artisan will understand that rigid body rotations are conveniently described by three Euler angles (φ,θ,ψ). Specifically, Euler angles (φ,θ,ψ) describe how body axes (Xb,Yb,Zb) originally aligned with the axes (X,Y,Z) of a coordinate system transform after three rotations are applied in a pre-established order. The magnitudes of Euler angles (φ,θ,ψ) define rotation of body axes (Xb,Yb,Zb) in the above-defined order. A skilled artisan will also be well versed in alternative rotation conventions and descriptions thereof. These will not be delved into further detail in this specification. For clarity and ease of explanation in the below teachings, we will sometimes use the angle θ with respect to a known axis in two-dimensional space to indicate the orientation of a sensor.
Preferably the location information of a candidate site is represented by (x,y,z) coordinates in three-dimensional Euclidean space, and the orientation of the sensing station is omni-directional, that is, unconstrained. Preferably the location information of a candidate site is represented by just (x,y) coordinates in two-dimensional Euclidean space, and the orientation of the sensing station with respect to an axis of the two-dimensional coordinate system is represented by the angle θ. Preferably the location information of a candidate site is represented by (x,y) in two-dimensional Euclidean space, and the orientation of the sensing station is omni-directional, that is, unconstrained.
Note that when we refer to an unconstrained orientation of a sensing station or characterize its orientation to be omni-directional above, that simply means that the sensing station is able to sense in all directions, irrespective of where it is ‘facing’. In other words, there is no front or back, or top or down, of the sensor. As will be apparent to those skilled in the art, a variety of such omni-directional sensors are commonplace in the industry, such as a 360° omni-directional or panoramic camera or an analogous microphone.
Similar to a candidate site, the location information of a target site may include its two-dimensional or three-dimensional coordinates in two-dimensional or three-dimensional Euclidean space of the system. The orientation information of a sensed station may include its three axes of orientation with respect to a given coordinate system. The orientation may be represented by rotation matrices Rx(∝) for rotation by angle ∝ around x-axis, Ry(β) for rotation by angle β around y-axis, Rz(γ) for rotation by angle γ around z-axis, or by Euler angles or still by any other rotation convention familiar to people of skill.
Preferably the location information of a target site is represented by (x,y,z) coordinates in three-dimensional Euclidean space, and the orientation of the sensed station is omni-directional, that is, unconstrained. Preferably the location information of a target site is represented just (x,y) coordinates in two-dimensional Euclidean space, and the orientation of the sensed station with respect to an axis of the two-dimensional coordinate system is represented by the angle θ. Preferably the location information of a target site is represented by (x,y) coordinates in two-dimensional Euclidean space, and the orientation of the sensed station is omni-directional, that is, unconstrained.
Similar to a sensing station, when we refer to an unconstrained orientation of a sensed station or characterize its orientation to be omni-directional above, that simply means that the sensed station is able to be sensed from all directions, irrespective of where it is ‘facing’. In other words, there is no front or back, or top or down, of the sensor. Again, as will be apparent to those skilled in the art, a variety of such omni-directional sensors are commonplace in the industry, such as an omni-directional radio transmitter with a dipole antenna.
Referring to
Still differently put, sensing region vk(p) represents the region of workspace 102 around candidate site p in which a sensing station can sense another sensed station. In case of the preferred embodiment depicted in
The invention further defines a sensing range and a sensing orientation as constraints that may apply to a given sensing station. These constraints are typical of the real world sensors available in the industry. For example, while a standard radio receiver can be an omni-directional sensing station with no sensing orientation or directional constraint, in the form of a parabolic dish however, a radio receiver can also be a directional antenna. In the example illustrated in
The invention further allows a given sensing station to have multiple sensors on it, each with its own sensing region defined above. Thus according to the foregoing formal definition of a sensing region, a sensing region vk(p) around a sensing station located at a candidate site p in workspace 102 represents the collection of target sites b in workspace 102 where bεvk(p) if a sensed station at site b is able to be sensed by sensor k of the sensing station despite obstructions 104 in workspace 102. Differently put, a sensing region corresponding to a given sensor on a sensing station when that sensing station is placed at a candidate site represents the collection of sites or locations in workspace 102 that can sensed by that sensor of the sensing station. Of course it is conceivable within the scope of the present invention to have a single complex sensor on a sensing station that has multiple sensing capabilities with multiple sensing regions according to the above definition.
Following directly from above, according to the present invention, a composite sensing region around a sensing station is defined as the collection of the individual sensing regions around that sensing station. As mentioned above, most likely but not necessarily, these individual sensing stations may be due to individual sensors on the sensing station. More rigorously, a composite sensing region v(p) of a sensing station is defined as the collection of all k sensing regions vk(p) when the sensing station is at a candidate site p in workspace 102.
Note that for clarity in
Recall that set X of target sites represents the collection of all points of interest or targets sites that are required to be observed. Recall also that the present invention allows for the placement of sensed stations at such target sites such that the sensed stations are able to be sensed by sensing stations placed at candidate sites. Also recall, in the present embodiment shown in
According to the invention, there is a set family comprising a set of ranges. Note, that following the standard practice of set theory, we are using the well-established term ‘range’ here to describe subsets of set X as will be further taught below. The term range from set-theory here is not be confused with the transmission range or reception range of a sensor. This unfortunate coincidence of reuse of the term in two different fields is unavoidable and any skilled artisan will be expected to understand the different notions of a ‘range’ as applied to set theory and sensors as obvious from the context in below teachings.
Each range in the set family of ranges corresponds to a given candidate site and represents the subset of target sites from set X that are able to be sensed by a given sensing station when that sensing station is placed at that candidate site. Recall from earlier teachings that a candidate site comprises the location information of sensing station in workspace 102, the type of sensor or sensing station deployed, and any other ancillary information about the sensing station. Thus each range in set family
corresponds to a given sensing station and represents a collection of target sites from set X that are able to be sensed by that sensing station when that sensing station is at the candidate site corresponding to that range in set family
.
The reader is encouraged to note, that the sensor deployment strategy offered by the present invention provides an effective mechanism to deploy a variety of different ‘types’ of sensors, a distinction over prior art. Thus it is sufficient for a range to be defined as per the above definition for each type of sensing station available in system 100, in our case omni-directional sensors with no range constraint such as sensor 106, directional sensors with no range constraint such as sensor 108 and directional and range constraint sensors such as sensor 110.
More formally, according to the invention, there is a set family of ranges ={R1, R2, . . . , Rm} whose union is the set X of all target sites, where Ri is the subset of set X of those target sites that are able to be sensed by that sensing station (or that type of sensing station as per above) when it is at a candidate site pi in workspace 102. From here onwards, we will generally drop the distinction between individual sensing stations and individual types of sensing stations to reduce repetition in the following explanation, and will only refer to the ranges being defined for each sensing station, with the knowledge that this implies defining the ranges for each type of sensing station. However as needed, we may distinguish the sensor types by their reference numerals 106, 108, 110 in the ensuing explanation.
The sensor deployment system 100 of }. Explained further, sensor deployment system 100 of the present invention determines a set of placement sites from the overall set of candidate sites {p1, p2, . . . , pm} for the placement of sensing stations 106, 108, 110 in workspace 102 to ensure coverage of the entirety of workspace 102 (excluding obstructions 104) such that any target site or point in workspace is sensed by at least one sensing station, by finding a minimum set-cover for the set system with the ground set as set X and ranges given by set family
. Note that the cardinality of set
is the same as the cardinality of the set of all candidate sites {p1, p2, . . . , pm}. In other words, there is a 1-to-1 correspondence between each candidate site, and what subset of set X (or the range), a sensing station at that candidate site can sense.
Those skilled in the art will understand that finding a minimum set-cover is an NP-hard problem. However a Greedy algorithm based solution is a popular approach to finding a near-optimal solution in polynomial time. Therefore, the minimum set-cover is preferably derived using the Greedy algorithm solution. Given sensor deployment system 100 of ={R1, R2, . . . , Rm} of ranges as explained above and the set X of all target sites, the Greedy solution can be implemented using the following pseudo-code:
It will be apparent to those skilled in the art how to implement or code the above popular algorithm using the appropriate data structures and other software programming constructs. Those details are well understood by skilled artisans and will not be delved into detail in this specification. For example, one approach is to store the target sites visible from a sensing station in the data structure associated with the candidate site where the placement of the sensing station is being considered, as explained earlier in reference to ancillary information associated with a candidate site. Based on the contents of this data structure, it is easy to define the ranges with respect to each candidate site and sensor, to implement the above algorithm in any programming language of choice.
Such a solution derived by the above Greedy algorithm representing a sensor deployment strategy for system 100 is illustrated in
This is because the Greedy algorithm in each iteration will choose the range with the largest cardinality (as shown in above pseudo-code), and hence will always favor the range corresponding to the most far-reaching or encompassing sensor, in our case sensor of type 106. Note also, that for clarity,
It will be obvious to those skilled in the art that a set-cover solution in computational geometry represents the minimum set of ranges that cover the entire ground set, in our case set X of all target sites. However such a solution is not unique in the sense that more than one solutions may exist that offer the same near-optimal deployment of available sensors. In other words, referring to our example and the solution offered in
In computational geometry, a sampling is an arrangement of points in the space chosen randomly, pseudo-randomly, along a regular grid, etc. Indeed through sampling a geometric problem is easily converted into a finite set system. Note that in our earlier example we were interested in observing all the sites or points within workspace 102 (excluding obstructions 104). However such is not always the case. In fact, very often it is desired to discretely sample a workspace such that there are a finite number of discrete points that comprise candidate sites {p1, p2, . . . , pm} and ground set X. Note that using the norms of computational geometry we are referring to our set X of all target sites in workspace 102 from
It should be noted that in the preferred embodiment explained above candidate sites {p1, p2, . . . , pm} and ground set X are overlapping, that is, a candidate site can also be target site or vice versa. However, in an alternative embodiment of the invention, candidate sites and target sites do not overlap. Such a situation is expected when there is a set of locations, such as walls or ceilings for sensing stations, and there are points or locations of interest on the floor or other parts of the building that are required to be observed.
Now we will look at a variation of the embodiment explained earlier with the added constraint that placement sites for a certain type of sensor can only be chosen from a set of sampled points or a placement region and that only a certain quantity of certain types of sensors are available. Such constraints are commonplace in real environments where cameras and other sensors are available in limited quantity and they can only be placed at appropriate locations in a workspace, such as walls and ceilings of a certain height, shape, construction, etc. Such a scenario is represented in
Let us further impose the constraint that five sensors of sensor type 106 are available, and 1 sensor each of types 110 and 112 are available. Indeed the capability to incorporate such constraints as to where potential candidate sites can be located and how many sensors of a given type can be used, represents one of the highly preferred embodiments of the present invention.
Based on a variation of the Greedy algorithm presented above,
Now we will look at the changes or variations to the Greedy algorithm presented above, that are required to enable this embodiment having constraints on the placement and quantities of sensors. Note that in as far as the constraint requiring the placement of the sensing stations at predetermined locations or placement regions, this constraint is easily satisfied by the selection of candidate sites {p1, p2, . . . , pm} in the first place. In other words, we will only pick candidate sites that satisfy the placement constraints of the problem i.e. candidate sites within placement regions 310, 312, 314 (see
As far as satisfying the constraint of predetermined quantities of various types of sensors, the Greedy algorithm above can be modified to include a ‘tally’ for each type of sensor. During each iteration of the algorithm shown, before choosing a range the algorithm will first check if the tally for the corresponding sensor is greater than zero. Once it picks the range with the largest cardinality, it will also decrement the tally of the corresponding sensor. If the tally reaches zero, the algorithm will stop choosing ranges corresponding to that sensor in subsequent iterations. Once all tallies have reached zero, the algorithm will terminate, whether or not a full cover has been computed. Note that this will be a ‘best-effort’ variation of the popular Greedy algorithm presented by the pseudo-code earlier. Such a best-effort algorithm will not guarantee that the computed solution will cover the entire ground set X, and may only provide a partial cover. Indeed the solution illustrated in
As taught above, preferably the minimum set-cover that forms the basis of the sensor deployment strategy of the present invention is based on the familiar Greedy algorithm. In the preferred embodiment, the minimum set-cover determined by the present invention is found in polynomial time. Those skilled in the art will know that Greedy algorithm gives an approximation ratio that is bounded by (1+log RL), where approximation ratio is defined as the size of the computed cover C divided by size of the optimal cover C* i.e. |C|/|C*|, and RL is the range set with the largest cardinality in set family . The Greedy algorithm is effective for applications where the size of set RL i.e. |RL| is a small fraction of the size of ground set X i.e. |X|.
An alternative algorithm is the one proposed by Bronnimann and Goodrich in “Almost optimal set covers in finite VC-dimension” (1995). Using this approach, given our set system Σ={X,} above having a VC-dimension of d and using the familiar Big-O notation, there is a polynomial-time algorithm for finding a set cover with size at most a factor O(d log dC*) from the optimal size C. Note that this bound does not explicitly depend on the cardinality of set X or the largest set in
. Preferably, the solution used to determine the minimum set-cover as taught by the present invention is of size at most a factor O(d log dC*) from the optimal size C. Still preferably, the VC-dimension d above is bounded by O(log h) where h represents the number of obstructions, such as those represented by reference numeral 104 in
Recall from earlier teachings that the instant invention allows a given sensing station to have multiple sensors on it, each with its own sensing region as defined above. Specifically recall that there are k sensing regions vk(p) around a sensing station located at a candidate site p and that there is a composite sensing region v(p) of the sensing station as a collection of all k sensing regions vk(p) when the sensing station is at candidate site p in the workspace.
Such a collection of individual sensing regions vk(p) can preferably be a union of the individual sensing regions vk(p) to form the resultant composite sensing region v(p) i.e. v(p)=∪v
In an alternative embodiment the above collection of individual sensing regions vk(p) can preferably be an intersection of the individual sensing regions vk(p) to form the resultant composite sensing region v(p) i.e. v(p)=∩v
Indeed in yet another preferred embodiment, the invention allows for a generic set operation to be performed on the individual sensing regions to form the resultant composite sensing region as may be required for a given application. Continuing our example above, if a third sensor on the sensing station is a microphone, then a positive confirmation of the presence of the short-range radio device may be obtained by the following definition of the composite visibility region of the sensing station: v(p)=((vradio ∪Vaudio)∩Vvisual)(p) i.e. either a radio or audio signal would suffice, as long as it is obtained with a visual confirmation. To conclude this discussion, and as previously mentioned, it is entirely conceivable within the scope of the instant invention to have a single complex sensor on a sensing station that has multiple sensing regions vk(p), for example, an Audio-Visual camera equipped with a lens and a microphone.
In a highly preferred embodiment, the present invention uses sensed stations located at the target sites.
According to the instant invention, sensed region or visibility region of a sensed station at a target site represents the collection of sites or locations in a workspace that reveal the sensed station to a sensing station when that sensing station is placed at a candidate site in the workspace and there is some overlap between the sensing region of the sensing station and sensed region of the sensed station. More rigorously, a sensed region μl(q) around a sensed station located at a target site qεX in a workspace represents the collection of sites c in the workspace such that the sensed station is able to be sensed by a sensing station at a candidate site p if site c is also in sensing region vk(p) of the sensing station despite the obstructions in the workspace.
Still differently put, a sensed region around a sensed station located at a target site q represents the collection of sites in the workspace that in case of any overlap with a sensing region of a sensing station location at a candidate site p will result in the sensed station being sensed by that sensing station. Of course, it follows directly from the above teachings that a composite sensed region μ(q) for a said sensed station at a target site q will be a collection of all individual l sensed regions μl(q) when the sensed station is at the target site qεX in the workspace. Indeed as in the case of sensing regions, a composite sensed region μ(q) around a sensed station as a collection of individual sensed regions μl(q), may be a union of individual sensed regions μl(q) i.e. μ(q)=∩μ
Similar to a sensing station, the individual l sensed regions μl(q) when the sensed station is at a target site qεX in the workspace may be as a result of the individual sensed regions of the various sensors present on the sensed station. The reader will observe that the teachings in reference to a composite sensing region v(p) of a sensing station and the various application scenarios, also easily extend to a composite sensed region μ(q) of the sensed station. In other words, as an example, it is easily conceived within the scope of the present invention to require a sensed station to either send a visual confirmation, or an audio and radio signals together to a sensing station. To satisfy these requirements, the composite sensed region μ(q) of a sensed station may be defined as: μ(q)=((μradio∩μaudio)∪μvisual)(q).
Preferably, sensed region μl(q) around each sensed station when said sensed station is at said target site qεX in a workspace is further defined such that a sensing station at candidate site p with sensing region vk(p) is able to sense the sensed station, if p is in sensed region μl(q) and q is in sensing region vk(p). Such an embodiment further tailors the application of instant invention to scenarios with more restrictive communication regimes, that is, where in order for a sensor to be sensed by another sensor, both sensors have to be within the sensing/sensed regions or fields of communication of each other, rather than merely having an overlap of their respective communication fields or radiation patterns.
Armed with the above definitions, let us return to
In a very highly preferred embodiment of the present invention, target sites are not merely locations or points in a workspace, but sensed stations or smart sensors, with their own sensed regions according to above explanation. This unique capability of the present invention, allows the sensor deployment based on minimum set-cover as taught above to be applied to a variety of interesting applications in a number of industry verticals that have requirements to observe and detect not just passive parts of a geography or ‘locations’ of interest, but rather monitor active devices with their own smart transmission capabilities and their own unique configuration and characteristics.
As taught earlier, candidate site(s) of the current invention comprise the location information of sensing station 504 in workspace 502, and any other ancillary information that may be needed to be associated with sensing station 504. Such ancillary information may include, but is not limited to, the type of the sensor, its configuration, orientation, sensing region, etc. In the preferred embodiment shown in
Similarly, system 500 also has sensed station 520 located at target site q at coordinates (x2,y2) with respect to the X,Y coordinate system shown, and has a sensed region 522 with an orientation constraint indicated by angle 524 as shown and a range constraint indicated by radius r2 of its conical sensed region 522. As per earlier embodiments, target site q represents the area of interest that is required to be observed by system 500 using sensing station 504.
Now let us place a further constraint on system 500 that sensor 504 can only be deployed within placement region 530 indicated in
Recall that in a preferred embodiment, sensed station 524 can be sensed by sensing station 504 if there is some overlap between their respective sensed and sensing regions. This overlap is indicated by reference numeral 512 in
Another preferred embodiment requires that the set of placement sites to be computed are such that each sensed station is able to be sensed by at least two sensing stations. This capability enables the invention to perform triangulation to determine the location of a sensor or sensed station. If each sensing station is covered by three sensing stations, then another technique called trilateration can be performed to determine the location of a sensor or sensed station. We refer to the capabilities of performing triangulation or trilateration as ‘localizability coverage’ of the present invention.
Those skilled in the art will understand the basic mechanism of triangulation for determining the location of a point by measuring angles to it from known points at the two ends of another fixed baseline. The point can then be fixed as the third point of a triangle with one known side and two known angles ∝ and β. In “Approximation Algorithms for Two Optimal Location Problems in Sensor Networks”, Efrat et al. of University of Arizona, Tucson (2005) proves that if there is a set-cover G1 then any target site in set X of target sites can be ‘two-guarded’ or sensed by two sensing stations by choosing one placement site from G1 and the second placement site from a second computed set-cover G2—albeit with an observation angle of α/2.
Such an approach can be easily implemented using the current invention by computing a set of placement sites or the first cover as in the main embodiment, and then performing a second pass by first removing the original set of placement sites from the available candidate sites {p1, p2, . . . , pm} and computing a second cover or set of placement sites to obtain the ‘two-guard’ solution as explained above. Such a two-guard solution will provide at least two placement sites that can sense each sensed station, and hence can be used to triangulate the position of any sensed station in the workspace (provided the observation angle is satisfactory for the requirements). Similarly, within the scope of the invention and using above techniques, one can compute a ‘three-guard’ solution of the target sites in set X to be able to trilaterate the position of any sensed station. Those skilled in the art will be familiar with the mechanism of trilateration using three known positions and will know that good trilateration is achieved when each sensed station is contained inside some triangle formed by three sensing stations.
Recall the earlier definition of a sensing region vk(p) around a sensing station at a candidate site p as the region in which the sensing station can sense a sensed station. Also recall the earlier definition of a sensed region μ(q) at a target site q around a sensed station as the region which if it intersects with a sensing region of a sensing station will result in the sensed station being sensed by the sensing station. We will refer to such coverage as ‘sensing coverage’. A highly preferred set of embodiments of the invention expand the above capabilities of sensing coverage to include the ability to communicate and not just sense. Consequently we will refer to such coverage as ‘network coverage’. Obviously network coverage implies sensing coverage.
Such communication can take a number of different forms including but not limited to sending and receiving messages that may contain just ‘pings’ or data payload, sending and receiving different types of electromagnetic radiation patterns to mean different things, sending and receiving different types or strengths of electrical signals to mean different things, sending and receiving different types or strengths of audio signals to encode different meanings and varying any characteristic of a physical signal to encode messages, etc.
Let us address the expanded definition of a sensing region of the current invention under network coverage more precisely. To enable network coverage, sensing region of a sensing station at a given candidate site represents the collection of sites or locations in the workspace where the placement of a sensed station will enable the sensing station to communicate with that sensed station. More rigorously, a sensing region vk(p) around a sensing station located at a candidate site p in a workspace represents the collection of target sites b in the workspace where target site bεvk(p) if the placement of a sensed station at site b will allow the sensing station to be able to communicate with the sensed station despite the obstructions in the workspace. Further, a composite sensing region v(p) of the sensing station is a collection of all such k sensing regions vk(p) when the sensing station is at candidate site p in the workspace. Still further, such a collection can be a union of all such k sensing regions vk(p) i.e. ∪v
Similarly, let us address the expanded definition of a sensed region of the current invention under network coverage more precisely. To enable network coverage, sensed region of a sensed station at a target site q in a workspace represents the collection of sites or locations in the workspace that in case of an overlap with a sensing region vk(p) of a sensing station at a candidate site p, will enable the sensed station to communicate with the sensing station. More rigorously, a sensed region μl(q) around a sensed station located at a target site qεX in a workspace represents the collection of target sites c in the workspace such that the sensed station is able to communicate with a sensing station at a candidate site p if c is in sensing region vk(p) of the sensing station, despite the obstructions in the workspace.
Preferably, sensed region μl(q) around each sensed station when said sensed station is at said target site qεX in a workspace is further restricted such that the sensed station is able to communicate with a sensing station at candidate site p with sensing region vk(p), only if p is in sensed region μl(q) and q is in sensing region vk(p). Such an embodiment further tailors the application of instant invention to scenarios with more restrictive communication regimes, that is, where in order for a sensor to communicate with another sensor, both sensors have to be within the sensing regions or fields of communication of each other, rather than merely having an overlap of their respective communication fields or radiation patterns.
Further, a composite sensed region μ(q) of the sensed station is a collection of all such l sensed regions μl(q) when the sensed station is at target site q in the workspace. Still further, such a collection can be a union of all such l sensed regions μl(q) i.e. ∪μl(q), an intersection i.e. ∪μl(q), or it can be based on a generic set operation on sensed regions μl(q). Obviously there are many applications of network coverage as enabled above in the real world. From monitoring pets with radio collars that need to communicate the identification number of the pet and location back to the owner, to babies with Radio Frequency Identification (RFID) bracelets, to toll stations such as bridges reading toll tags in vehicles, to name a few.
In the preferred embodiment of the invention the sensing stations and the sensed stations of the present invention are wireless devices that operate substantially in the 60 GHz range. Those skilled in the art will agree that 60 GHz frequency range is poised to become the next big frequency in the world of wireless devices, with both short-range and wider area applications. The frequency is part of the ‘V-Band’ frequencies in the United States and is considered among the millimeter radio wave (mmWave) bands. Its applications will include broad range of new products and services, including high-speed, point-to-point wireless local area networks and broadband internet access. High Definition Wireless (WirelessHD) is another recent technology that operates substantially near the 60 GHz range. A key characteristic of this frequency range is that its highly directional, ‘pencil-beam’ signal characteristics permits different systems to operate close to one another without causing interference. The upcoming Wi-Fi standard IEEE 802.11ad is also slated to run in this frequency range.
In another preferred embodiment of the instant invention, the workspace is a video, whether realtime or near-realtime streaming video or pre-recorded footage. In this interesting embodiment the applications include finding a scene in the video that would cover a desired place, such as a geographical location in an urban or sub-urban environment, a building, or a specific room in the building. Examples of such an application are in video editing where the director and video editor are interested in ensuring that a certain part of the set, such as a house or the living room of the house, is adequately covered in the final edited footage that was originally taken by a number of different cameras from different locations. Alternatively, a criminal investigation may be concerned with ensuring that a certain event that had transpired is as fully covered by the available footage from passerbys and security cameras on the surrounding building as possible. The present invention provides such a capability by mapping the sensing stations to be the cameras with their associated timelines, the workspace to be the entire footage, and the points of interest to be the place, event or the scene that needs to be covered.
In a similar embodiment of the present invention, the sensing stations are persons, sensed stations are other persons, the workspace is a given geographical area while the candidate and target sites comprise the location coordinates in the geographical area. In a related embodiment, the sensing stations or sensed stations are other objects of interest, and not necessarily human beings. Of course, many applications fitting such embodiments are easily conceived. An example use-case for such an embodiment is ensuring that celebrities (sensed stations) at the Academy Awards ceremony or The Oscars at Dolby Theater in Hollywood (workspace) are adequately covered by NBC videographers (sensing stations). Another example could be vehicles in a parking lot or a transportation hub that need to be tracked by sensors, etc.
In yet another set of highly preferred embodiments of the present invention, a social graph is treated as the workspace. With the ubiquitous presence of social networking communities such as Facebook, LinkedIn, Google+, MySpace, Instagram, Tumblr, YouTube the notion of a social graph is ever more important, at a personal level for the user, at a commercial level for marketers of products and services, and even for law enforcement agencies. To explain these embodiments, let us look at sensor deployment system 600 illustrated in
Sensing stations are all other persons in the social graph shown by blank circles 606 in
The significance of such a capability is profound from a marketing and outreach perspective. One can easily conceive of marketing and outreach campaigns that are designed to reach certain demographics or subsets of the social graph of a community. An exemplary situation will be an election candidate reaching out to a certain subset of the population. Note that in an alternative but similar embodiment, the ground set X may not comprise people, but rather products. For example, in marketing analysis a marketer is interested in knowing what customers are using what products. In such a situation an alternative graph comprising people and the products they are using may be constructed and used as the workspace for the present embodiment, and the desired analytical objectives achieved.
Indeed the definitions of sensing and sensed regions are applicable to the above embodiments having a social graph as the workspace. While in the example above, sensing region of sensing stations shown by circles 608 in
One can also easily extend the notions of sensing and network coverages taught above to the present embodiments using a social graph. While a sensing coverage in a social graph implies the knowledge or the existence of person within the sensing or sensed regions as per above definitions, a network coverage would allow communication between popular persons and their friends, and their friends and so on.
The methods of the present invention further delineate the steps required to execute the sensor deployment strategy of the present invention taught above. The methods provide the steps for determining the placement sites from a set of candidate sites {p1, p2, . . . , pm} for sensing stations in a workspace, by first providing sensed stations at target sites in the workspace, and representing all such target sites by set X. They further provide zero or more obstructions in the workspace, and then provide one or more sensing regions vk(p) around each sensing station when the sensing station is at a candidate site p in the workspace. Sensing region vk(p) is a collection of all sites b in the workspace such that the sensing station at candidate site p is able to sense the sensed station at site b, despite the provided obstructions.
In related embodiments, the invention further extends the definition of sensing region vk(p) beyond sensing coverage to include communication and hence provide network coverage. Consequently in such embodiments, sensing region vk(p) is defined as a collection of all sites b in the workspace such that the sensing station at candidate site p is able to communicate with the sensed station at site b, despite the provided obstructions.
The methods further provide a sensing range and a sensing orientation to constrain the sensing region vk(p) of each sensing station, and then provide a composite sensing region v(p) of a sensing station to be the collection of the individual k sensing regions vk(p) when the sensing station is at candidate site p in the workspace. Furthermore, set family ={R1, R2, . . . , Rm} whose union is the target set X is created, such that a sensing station at a candidate site pi in the workspace is able to sense each sensed station at the target sites in set Ri belonging to set family
. Then the step to choose the placement sites from the set of candidate sites {p1, p2, . . . , pm} for the sensing stations in the workspace is performed by computing a minimum set-cover for set system Σ={X,
}. As taught earlier, the minimum set-cover can be computed using the popular Greedy algorithm, or a variation thereof, or any other suitable algorithm appropriate for the application at hand.
The methods of the present invention extend the above steps by providing a sensed region μl(q) around each sensed station when the sensed station is at a target site qεX in the workspace, and then setting the sensed region μl(q) to be a collection of all sites c in the workspace such that the sensed station is able to be sensed by a sensing station at a candidate site p in the workspace, provided c is also in the sensing region vk(p) of the sensing station. In related embodiments, the invention further extends the definition of sensed region μl(q) beyond sensing coverage to include communication and hence provide network coverage. Consequently in such embodiments, sensed region μl(q) is a collection of all sites c in the workspace such that the sensed station is able to communicate with the sensing station, provided c is also in the sensing region vk(p) of the sensing station at candidate site p in the workspace.
Further a sensed range and a sensed orientation is provided to constrain sensed region μl(q) of the sensed stations. Then a composite sensed region μ(q) for each sensed station is provided as a collection of all individual l sensed regions μl(q) when the sensed station is at target site qεX in the workspace.
In variations of above embodiments, the methods restrict the definition of sensing region vk(p) and sensed region μl(q) such that a sensing station at a candidate site p in the workspace is able to sense a sensed station at target site q in the workspace or communicate with it, only if p is in sensed region μl(q) of the sensed station and q is in sensing region vk(p) of the sensing station.
In view of the above teaching, a person skilled in the art will recognize that the methods of present invention can be embodied in many different ways in addition to those described without departing from the principles of the invention. Therefore, the scope of the invention should be judged in view of the appended claims and their legal equivalents.