JP2004509389A - Method and apparatus for ground inference with distributed embedded processing elements - Google Patents

Abstract

Methods and apparatus are provided for calculating characteristics of the physical environment using a plurality of agents forming a distributed network embedded in the environment. The method includes determining an initiating agent 200, transmitting a signal including an accumulated cost value to an adjacent agent 202, and transmitting the signal at each adjacent agent to enhance the accumulated cost value with local information 204. Processing. If more than one signal is received, determine which has the best accumulated cost value to generate a new signal 206, then treat the neighbor agent as the initiating agent 208, and The new signal is sent to 208 and the memory 210 retains the best enhanced cost value. The method further includes determining the path using a shortest path calculation, using a double gradient to align the agents on one path between the two reference agents, Discovering and converging the agent.

Description

[0001]
[Statement of rights to be attributed to the federal]
The present invention relates to research conducted within a contract with the United States Government for a DARPA ITO Agreement No. N66001-99-C-8514, "Pharmone Robotics", The invention has certain rights.
[Priority claim]
No. 60 / 217,232, filed in the United States on July 10, 2000, entitled "Emergent Movement Control for Large Group of Robots", and No. 60 / 217,226, filed in the United States on July 10, 2000, entitled "Terrain Reasoning with Distributed Embedded Processing Elements", entitled "Terrain Reasoning with Distributed Embedded Processing Elements."
Claim the benefits of priority.
[0002]
TECHNICAL FIELD OF THE INVENTION
The invention relates to the field of calculating properties of a physical environment. More specifically, this disclosure presents a method for calculating characteristics of a physical environment using multiple agents forming a distributed network embedded within the environment.
[0003]
Problems to be solved by the prior art and the invention
When it comes to path planning and approaches to ground analysis, over the past few decades, there has been interest in single or parallel processor solutions that operate on internal maps containing ground features. Traditionally, ground analysis has involved one of two forms: 1) a single mobile agent (eg, a robot) or a small number of agents that collect ground information by flying over; And / or use pre-built digital maps to assist in path planning or other ground related decisions. 2) As pre-built digital maps useful for geometric calculations, ground / environment databases (simulations as below) (For example, a simulated ground vehicle or a simulated enemy in a computer game).
[0004]
Typical calculations include determining the shortest path to a destination, determining a location with high visibility on the ground, and determining a hidden route from a known enemy location. In such traditional approaches to route planning and ground logic, relevant data is taken from the real world to create a machine-readable map. Ground / environment logic based on local sensing provides very limited access to information about the environment.
[0005]
Pre-building a map of the environment with solutions from the natural world provides much more information and can be updated with information from the robot's own sensors. Such maps are usually represented by an array of ground characteristics, including traversability and blocked routes, which can be translated into a corresponding array representing the difficulty and cost of traveling across a particular ground path. .
[0006]
The cost array can then be analyzed to obtain a new map showing the minimum cost path between designated points on the map. In this traditional approach, the separate and distinct steps of sensing the environmental characteristics of the area, sending data to a central point, and generating a map, are performed when all minimum cost path calculations are performed. Must be done before As a result, in situations where the environment is changing rapidly or sensing and map generation cannot be performed in a timely manner, it is often difficult to determine the minimum cost path incorporating the latest information. is there.
[0007]
Current means of mapping space include the use of several powerful mobile robots that navigate the environment in a search pattern and send raw or partially processed data back to a central processing resource. Included. The central resource edits the map and then performs operations on the map. Dramatic changes to the environment are not captured until the robot repeats its search.
[0008]
In some communication networks, calculations are performed to determine the lowest cost path for communication purposes (eg, to minimize delay due to transmission energy or signal retransmission). These approaches are particularly relevant to telecommunications issues, and they do not attempt to derive a correspondence between communication efficiency and environmental characteristics that might affect mobility. . Information relevant to some of these approaches can be found in U.S. Patent Nos. 5,561,790 and 5,233,604.
[0009]
Thus, a number of small, inexpensive agents that require only local communication with adjacent agents are used to provide a means for determining the optimal path over or through an area of an environment. It would be desirable to provide. The non-local characteristics accumulated along one path are signals transmitted across the path by making a series of hops through multiple agents in a manner that provides a local representation for each agent. Would. Information at the agent could be updated immediately by sending another signal along the path.
[0010]
[Means for Solving the Invention]
The invention disclosed herein provides a method and apparatus for calculating characteristics of a physical environment using a plurality of agents located within the physical environment. Each deployed agent includes a communication capability, a processor coupled to the communication capability, and a memory connected to the processor. The plurality of agents form a distributed representation of the non-local characteristics of the physical environment by utilizing the physical characteristics of the inter-agent communication.
[0011]
Triggering at least one initiating agent; sending a signal from the initiating agent to an agent adjacent to the initiating agent; communicating between the initiating agent and each adjacent agent Processing signals at each of said neighboring agents according to a cost value based on local characteristics along respective paths between; selecting a new signal to transmit from the received signals; Treating an agent as a triggering agent of the processed signal and repeating the transmitting and processing steps; at least one agent representing information about a non-local property of a physical environment based on the received signal. The storage cost value includes the step of locally retained.
[0012]
Thus, a signal propagates from the initiating agent to each of the plurality of agents, and the accumulated cost value is stored at each agent to provide a measure of quality for the best path from the agent to the initiating agent. You.
[0013]
The cost value is expressed as a simple hop count, and the signal propagation may be repeated from the initiating agent to multiple agents at each hop. The accumulated cost value is typically used to determine a new signal to be transmitted from the received signal.
[0014]
In addition, the cost value is retained from past signals and can be compared to the cost value from current signals to determine the cost value of processing the signal. The signal is typically passed along the lines of a visual path, which provides a general expression of the traversability along one path.
[0015]
The agent may also be augmented by providing the agent with a means for preserving the direction in which a particular signal has been received over one communication path, and the difference between the two agents The cost value for a path can be stored with the direction corresponding to the path with the best cost.
[0016]
Further, when the cost value is stored in the agent's memory, the time can be recorded, and the most recently stored cost value is specified to be used in processing the signal. And the cost value is removed from memory after a predetermined time has elapsed. The cost value can also be ranked from best to worst over time, and the best cost value is used in processing the signal.
[0017]
In this case, when the time value of the cost value specified to be used in processing the signal expires and the cost value is removed from memory, the next best cost A value is specified to be used in processing the signal.
[0018]
In another embodiment of the invention, the agents can be configured to perform an alignment, wherein two agents select from among the plurality of agents to act as reference agents. Generating a cost gradient comprising a vector representing an optimal cost path between agents among the plurality of agents and a cache-based agent; the vectors are summed at each agent to create a motion vector for the agent. And the agent is moved along the motion vector to align the agent along one path between the reference agents.
[0019]
In the aligning step, for example, a trigger can be pulled upon occurrence of a triggering event such as detection of an intruder. Once the agents are aligned along one path, they are converged to one choke point by selecting a convergence agent and prompting other agents to head towards the convergence agent. Is possible, resulting in local clustering of agents near the convergence agent.
[0020]
The agents can also incorporate sensors connected to the processors to provide additional information to the processors for generation of the cost value. For example, a smoke detector can be used to factor the extent of smoke in an area into the cost value along one path.
[0021]
Agents designed for use with the present invention can incorporate the logic used to implement the present invention in the form of software, or the logic can be encoded in hardware.
The invention relates to the field of calculating properties of a physical environment.
The following description is presented to enable one of ordinary skill in the art to make, use, and incorporate the invention in the context of a particular application. Various modifications as well as various uses in different applications will be readily apparent to those skilled in the art, and the general principles defined herein will apply to a wide range of embodiments. It is possible.
[0022]
Accordingly, the present invention is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein. Should be tolerated.
[0023]
To provide a reference work frame, a description of terms used in the description and claims is first provided as a central source for the reader. A brief introduction is then provided in the form of an illustrative description of the invention to provide a conceptual understanding prior to elaborating its specific details.
[0024]
[Glossary]
Before describing the specific details of the present invention, it is instructive to provide a centralized listing of various terms used in the specification and claims.
[0025]
Agent-As described herein, the term agent, in its most general sense, refers to a unit that includes a processor, communication capabilities, and memory. The agent also includes one or more means for movement and means for detecting its orientation relative to other objects in the agent placement system (compass, directional transmitter / receiver, etc.). , A distance measuring device (such as a laser, infrared, ultrasonic, or radar range finder) and one or more sensors.
[0026]
In even more straightforward terms, an agent can be, for example, a robot with an infrared transmitter / receiver and a set of wheels, where the infrared transmitter / receiver is connected to the robot and other robots. Used to measure the distance and direction to the robot. In more complex situations, the same robot may also have antennas to transmit and receive wireless communications, sensors to sense various conditions in the environment (temperature, visibility, etc.), flight, Alternative means of movement, such as wings and propellers for the vehicle, may be included.
[0027]
Choke point-In certain situations, it may be desirable to move a group of agents toward a point. For example, if an intruder is moving around in one area, multiple agents can be sent in the vicinity of the intruder. On the other hand, in areas with narrow corridors, it may be desirable to restrict the passage of intruders.
[0028]
Thus, an agent may be sent near a corridor along the intruder's path to limit the intruder's movement. Thus, a choke point is generally a convergence point of a group of agents around one selected agent, blocking a passage or placing a group of agents densely at one selected locality. Can be used to
[0029]
Communication Capability—The term communication capability as used herein generally refers to a local communication mechanism that facilitates communication between an agent and its local neighbors. Any communication mechanism applicable to a particular environment may be used.
[0030]
For example, in a clean atmosphere, a directional infrared communication mechanism may be particularly useful because of its directivity. On the other hand, in a water or liquid environment, sonic communication mechanisms may be more useful.
[0031]
Cost value-The term cost value refers to a measure of one or more physical characteristics along a path between agents. For example, in a very simple case, the cost value can be a measure of whether one agent was able to signal another agent, in which case the hop count would be the cost value. Could function as
[0032]
On the other hand, more complex cost values may be developed (for example) as a weighted sum of several physical properties, including, but not limited to, signal transmission quality, temperature, wind speed, Includes the angle of the path between the agents. Many cost value schemes can be deployed as needed for a particular implementation.
[0033]
The term local cost value is intended to refer to a measure of local environmental conditions along the communication path between two agents. The term accumulated cost value is intended to refer to a measure of non-local environmental conditions along the path between two or more agents. Although the terms local cost value and accumulated cost value are defined herein, the exact application of this term can also be inferred from its use.
[0034]
Also, the cost value generated at each of the plurality of agents responsive to the signal transmitted by the initiating agent forms a cost gradient pattern across the plurality of agents. Combined with the directional information, the cost gradient pattern has many features of the oriented graph, and many calculations performed using the oriented graph apply to the plurality of agents. It is possible to do.
[0035]
Local-Agents that are within direct (usually line-of-sight) communication and the sensing range of an agent are considered to be local and are called its neighbors. The number of local (adjacent) agents for one arbitrary agent may vary depending on the number of agents present in one arbitrary area. It is desirable to use signals only between local agents to conserve transmission power.
[0036]
For the case where it is used with respect to the properties of the environment, the properties that are local to one agent are the properties that can be sensed by said agent. Non-local physical / environmental properties for an agent generally refer to those physical / environmental properties that the agent cannot directly sense.
[0037]
Means for Providing Mobility-To provide mobility that needs to move around in the environment (assuming movement is desirable), agents need something to facilitate mobility. I do. Any known controllable movement mechanism can be used, as appropriate for the environment in which the agent is to be used.
[0038]
For example, on the ground, agents would typically have legs or wheels, while in the air they would have propellers and wings. On the other hand, in a liquid environment, the agent would be equipped with propellers and fins, and in space, the agent would be equipped with a jet or rocket. Means for providing mobility are coupled to and controlled by the processor.
[0039]
Sensors-In addition to means for detecting distance and direction from other agents, agents can include sensors associated with their status as well as sensors coupled to their processors to collect information about their environment. It is. Examples of sensors that can be used in the agents of the present invention include, but are not limited to, smoke detectors, audio detectors, visual sensors, seismic detectors, heat / infrared detectors, An article sensing device and a satellite based positioning system (GPS) sensor are included.
[0040]
Sensor information can be incorporated into the cost value along one path, or can be used to trigger various behaviors of the plurality of agents, such as convergence to a choke point .
[0041]
[Introduction]
The present invention generally provides a means for using a plurality of agents distributed in an area to perform terrain calculations on the area. Properties can be determined by creating a pattern of communication between the nearest neighbors, where the source agent sends a signal to each of its neighbors that includes the accumulated cost value. Send
[0042]
Each adjacent agent receives the signal from the source and modifies the accumulated cost value based on the physical characteristics between the source and the respective adjacent agent. And sends a new signal to its neighboring agent incorporating the modified cost value.
[0043]
As the signal propagates through multiple agents, a single agent can receive a signal having an accumulated cost value from two or more of its neighbors. In some particular embodiments, a single agent determines the best accumulated cost value among those received, and further modifies the best accumulated cost value into a new signal.
[0044]
Alternatively, the agent can modify all of the accumulated cost values to determine the best accumulated cost value and incorporate it into the new signal. The accumulated cost value represents a non-local characteristic of the environment (eg, a characteristic of the physical environment that cannot be directly sensed by the receiving agent). After modification, the agent saves the best accumulated cost value and generates a new signal to send to its neighbors.
[0045]
As a clearer example, multiple agents could be used to conduct a search for fires (or other objects such as mines, intruders, etc.). If an agent detects a fire while the user is waiting at a specific location, a signal from the detection agent is propagated throughout the plurality of agents and transmitted to the user.
[0046]
The user is provided with an optimal path to the fire (e.g., the fastest, most secure, etc.) since multiple paths with accumulated cost value are useful anywhere in the multiple agents. ). The best path can easily be designated as the path along which the signal has taken the fewest hops along the route to the user (as an estimate of the distance). On the other hand, if the best path is derived from a sensor mounted on the agent along the path, it may be a function of more complicated information.
[0047]
Generally, once a triggering agent begins to propagate a signal across multiple agents, the signal is modified at each signal pop between the triggering agent and each receiving agent. The modifications have the purpose of incorporating information about the characteristics of the physical environment across the latest conveyed hops, and can be as simple as updating the hop count, or as complex as adding sensor information or local computation results. sell.
[0048]
The information about the physical environment along each path can easily provide a measure of the quality of the path and provide a separable part to determine the unique characteristics at each hop in the path. It is possible to include. In any case, the stored signal specifically represents the quality of the best path. Propagation of the signal between the initiating agent and the receiving agent is basically performed by the initiating agent and the receiving agent at each step to provide a local overview of the non-local characteristics along the path. Is a distributed computation that occurs over the entire path between
[0049]
The level of detail information about environmental characteristics along a path may vary widely depending on the particular implementation. For example, one of the easiest ways to measure the environmental properties is to count the number of hops between the initiating agent and the receiving agent along the path of a successful transmission between two agents. Is to take advantage of the presence or absence of the ability to transmit signals.
[0050]
A slightly more complicated method is to use the strength of the signal received over one hop as a measure of the quality of the bus along the hop. This provides a more accurate way of determining the distance. Still more complex methods are to use sensors mounted on the agent to determine the physical environmental properties and to incorporate those properties into the signal.
[0051]
In this sense, the stored cost value embedded in the signal may serve as a weighted sum of the properties across the transmission path. One agent may receive a plurality of signals on the route from the starting agent to the receiving agent. In this case, the agent will select the signal representing the most suitable path to retransmit to its neighbors.
[0052]
In essence, the measure of the quality of the path along the hop between one agent and another agent can be considered as cost value, where the cost is low The lower, the better the path. If the quality of a path is easily determined from the hop count to that point, the cost value along the most recent hop is a binary on / off indicating whether the signal was able to cross the path. Off function. The accumulated cost values of multiple passes are compared to determine the best pass.
[0053]
Again, in a simple hop count embodiment, for each hop for which the signal is transmitted successfully, the path with the lowest accumulated cost value is considered optimal (the shortest or lowest hop, The accumulated cost value is increased by one (assuming the path is most desirable). The concept of cost value can be extended to other embodiments that are used with other sensors, where the accumulated cost value is more complex than many aspects of the environment are potentially embedded. It may be a good measure. In a volatile environment, the cost value may be designated to expire after a certain amount of time, possibly replacing the next best older (historical) cost value.
[0054]
In addition to simply determining the quality of the path between two agents, two agents may be selected and transmission of signals across the plurality of agents may be initiated simultaneously. Preferably, each of the two agents transmits signals in a unique and identifiable manner (eg, on different frequencies, with different signs, etc.). Each of the plurality of agents receiving the signal develops a vector representing the quality and direction of the best path for each of the two agents, and to line up on the path between the two agents, Move according to the two vectors.
[0055]
The selection of the two agents can be made explicitly by the user, or when a triggering event occurs, propagate the message from one agent along the path to a second (predetermined) agent. It can be done by doing. Next, another agent is selected from the lined-up agents to act as a convergence point for another agent, and the agents lined up on the path go to the convergence point You will be prompted to: The usefulness of this behavior is illustrated in the case where a fire is detected in one area of a building.
[0056]
The agent detecting the fire signals other agents to find a narrow corridor near the fire. Once a narrow corridor is discovered (perhaps by asking a nearby agent to determine what is facing the greatest motion limitation), pass through the corridor to another agent on the other side of the corridor The path is determined. Next, another agent along the path is selected to send a signal, while the other agents move to line up along the path. The agents in the narrow corridor then signal a flame shut-off agent along the path to move to the corridor, where they shut off the fire from the rest of the building In that attempt, a limited amount of fire-blocking chemicals can be applied.
[0057]
Now that the general nature of the invention has been described, the specificity of the invention will be described with reference to the drawings.
[0058]
BEST MODE FOR CARRYING OUT THE INVENTION
First, a description and illustration of one exemplary agent used by the present invention is presented. Then, use the agent to measure the best path path from the initiating agent and another arbitrarily selected agent, and to help other agents in converging on the choke point. The above methods of using two agents as reference agents will be discussed with reference to some flowcharts. Finally, an illustration is presented to graphically illustrate various aspects of agent placement.
[0059]
A typical agent used in accordance with the present invention is illustrated in FIG. As shown, the agent 100 includes a communication capability 102 for transmitting and receiving signals, shown as an antenna. Internally, the agent 100 has a memory for processing received signals and data about the environment, as well as controlling the means for providing mobility, such as a set of trucks 104 or wheels 106. Processor included.
[0060]
The agent 100 also includes one or more sensors, which are non-limiting examples shown, including a metal detector 108, a GPS receiver 110, an infrared detector 112, and a gas chromatography 114. May be included. The agent 100 may be adapted for use within a particular environment or may be adapted to detect particular characteristics of the environment. The agent 100 is used to calculate properties of the physical environment and develop a distributed representation of the non-local properties of the physical environment.
[0061]
An agent may be such that each agent 100 is within communication range of at least one other agent 100 and an agent that is within direct (non-relaying) communication of any agent with any locally adjacent agent of that agent. It is defined as 100. The plurality of agents 100 form a distributed network of communication processors as a whole.
[0062]
Preferably, the agents 100 are arranged such that they are substantially evenly spaced throughout the desired area, maximizing its ability to effectively cover the area. The placement of the agent can be by any means desired for the particular environment.
[0063]
For example, if the agent is mobile, various algorithms may be used to ensure that each agent 100 is moved to a location equidistant from the other agents 100. If the agents are not mobile, they may be dispersed in a semi-random manner (eg, drop them out of the aircraft), or they may be positioned more carefully. There is also a hybrid travel situation map, where only a portion of a group of agents 100 can travel.
[0064]
The goal of the present invention is to determine the optimal path from one initiating agent to any one of the agents. FIG. 2 is a flow diagram of a method for determining the optimal path, beginning with the step of determining at least one initiating agent 200, wherein an agent begins transmitting signals propagated through the plurality of agents. The initiating agent can be explicitly determined by a command from the user of the agent.
[0065]
For example, the user can use a remote control such as a laser pointer or wireless communication device to select one agent from which to begin propagating the signal. Alternatively, the initiator can be determined implicitly by one event. For example, the agent includes a sensor such as a smoke detector, and when detecting smoke by the agent, the detection agent may be designated as the starting agent. The exact mechanism by which one agent is established as the initiating agent is not particularly important.
[0066]
After the initiating agent has been determined, the initiating agent sends a signal to neighboring agents in a transmitting step 202. As mentioned earlier, an adjacent agent is considered to be an agent within the communication range of any given agent. Generally, each one of the plurality of agents will have at least one adjacent agent.
[0067]
The signal is then received at the neighboring agent, and in a signal processing step 204 according to the cost value associated with the respective path between the initiating agent and the neighboring agent. Be processed. The cost value may be based on transmission-related local characteristics along the path, such as the hop count along the path, or the quality of the transmission provided by the path (in a simple case, , The cost value may be based on whether the agent can communicate along the line of the visual communication path).
[0068]
Alternatively, the cost value may be based on local characteristics of the environment near the agent, such as temperature, humidity, degree of smoke, and the like. In essence, any local characteristics of the environment are incorporated into the fixed cost value. Further, it is possible to use a weighted combination of multiple properties.
[0069]
Next, each agent receiving the signal determines a new signal to be transmitted based on the received signal. The determining step 206 can be performed either before or after the signal processing step 204, depending on the particular implementation. The decision on the new signal to send is generally based on a comparison of the cost value of the signal received from the agent's neighbors.
[0070]
For example, if the goal was to determine the lowest hop count path from the initiating agent in preparation for propagating the signal, then the agent along the path would determine the signal with the lowest hop count. Then, the signals received from the adjacent ones are compared. The agent will then increase the hop count and send a new signal to the neighbor indicating the increased hop count.
[0071]
After determining the new signal to send, the agent essentially becomes like a new initiating agent in process step 208, and sends a new signal to its neighbors. These steps are repeated to propagate one signal across the plurality of agents.
[0072]
As the signal is propagated, the accumulated cost value is held in memory at each agent in a holding step 210. The accumulated cost value provides, at each agent, a local overview of non-local features along one path towards the initiating agent and also provides a mechanism to prevent back propagation of one signal I do. For example, in many configurations, agents can broadcast in all directions. Thus, after the first agent sends a signal representing its current cost value, the neighboring agent receives the signal and sends a new signal representing the cost value at the neighboring agent. Will be.
[0073]
Since the transmission from the adjacent agent is omni-directional, the new signal will be received by the first agent. Assuming that the cost value of the adjacent agent (which is derived from that of the first agent is less desirable than that of the first agent, the first agent It does not hold and does not make use of the cost value of the neighboring agent, so the propagation of signals from the initiating agent is generally directed away from the initiating agent.
[0074]
The accumulated cost value at any given agent represents an accumulated measure of quality for the path that the signal traversed along the route from the initiating agent. In this way, the accumulated cost value acts as a local representation (at the agent) of a non-local characteristic (to the agent) along the path between the initiating agent and any agent.
[0075]
In some cases, it may be desirable to compare the retained cost value with the received cost value from one signal to determine which to use in determining a new signal to transmit from the received signal . In this way, the best accumulated cost value received at each agent is retained, and each time a signal is passed to the agent, the received accumulated cost value is determined by the retained cost value to determine which is higher. It is possible to compare with value. For this to happen, it is assumed that the same initiating agent periodically retransmits the signal to be propagated across multiple agents.
[0076]
This may be desirable, for example, as part of an update process to continually update the accumulated cost value at each agent to track changes in the area covered by the plurality of agents. .
[0077]
The signal processing step 204 may also include the use of a time stamp to provide a mechanism for the cost value to expire over time. As cost values are generated and held in memory, they may be time stamped. The most recently timed and kept cost value can be used in processing the signal.
[0078]
However, after a predetermined period of time, the stored cost value is removed from memory. By using cost value expiration, cost value information is retained over a period of time, but it also expires after a certain amount of time has passed and the system One advantage is provided by ensuring that information that is no longer used is flushed.
[0079]
In addition to simply storing the final cost value in memory, multiple cost values are stored in memory in a stack-like configuration, with the cost values ranked (or accessed) in a certain order. The order of the cost values can be used, for example, from highest to lowest or vice versa, or some other ranking criterion.
[0080]
In this scenario, the newly received cost value is also recorded at that time and stored in memory. When the current cost value expires, the next best cost value stored in memory is used until it expires, and so on as long as the cost value remains in memory. . This mechanism allows the cost value stored in memory to follow the quasi-decay behavior before being erased.
[0081]
The agent can be equipped with means for determining the direction in which the signal was received. This can be achieved by using a directional communication mechanism such as directional infrared transceivers or directional radio receivers. An agent so equipped can store both direction and cost value for a single pass to an adjacent agent. To determine the optimal path to the initiating agent, rather than searching by neighbors for the one with the best cost value, each agent can save the direction to the next agent along the path.
[0082]
The direction in which the signal was received can also be factored into the cost value. For example, a path introduced to a particular destination may be less desirable than one introduced to another destination. Thus, the direction can be factored into the cost value to either increase or decrease the cost of a path in a certain direction. The agent 100 can also incorporate a mechanism for determining position and direction into the world coordinate system, such as a satellite positioning system (GPS) receiver and a compass to help facilitate direction determination.
[0083]
It is also possible to use multiple initiators, and the paths for each can be compared to determine which is better. For example, if agents are dispersed throughout an area known to have several mines, the signal may be detected based on the detection of mines near one or more agents. It may be desirable to trigger the start of propagation. Each agent can then initiate the propagation of a signal having a cost value.
[0084]
At each agent, the cost value of the signal initiated from each of the initiating agents can be compared to determine not only the optimal path to each, but also the overall optimal path. As a result, even if a plurality of mines may exist in the agent, it is possible to determine the optimum path to any one of the mines. In this case, it may be desirable for each agent in the plurality of agents to use a unique identifier to indicate a cost value along a path to each initiating agent.
[0085]
The cost value of the paths to each initiating agent can then be compared at each agent to determine the best path among them. The cost values are passed individually to each path to each initiating agent, or the cost values are compared and only the best is passed. In other words, each agent in the plurality of agents may receive and also retain a cost value for the path of each initiating agent, or may retain only the best cost value received. It is possible.
[0086]
In addition to easily identifying the path from an arbitrarily selected agent to an initiating agent, it may be desirable to utilize a path between two agents. For example, it may be desirable to have agents line up along one path between two reference agents. A flow diagram illustrating the steps used to line up agents along one path is illustrated in FIG.
[0087]
First, in the reference agent determination step 300, it is necessary to determine two agents from among a plurality of agents to act as reference agents. Next, one signal is transmitted from each of the reference agents. The transmission of the signal and the generation of the path vector are generally shown as a cost gradient generation step 302. The path vector essentially creates a cost value gradient over the area covered by the agents and towards the initiating agent.
[0088]
The path vector is therefore used in a motion vector generation step 306 to create a motion vector. In general, each motion vector represents the sum of the path vectors, although mechanisms other than simply producing a sum can be used. Agents are prompted to move along the continuously updated motion vector until they are positioned along one path between the two reference agents in an agent convergence step 308.
[0089]
The closer the agent moves to the path between the two reference agents, the smaller the motion vector is and the closer the path vector component is to the opposite side. Regarding the designation of the initiating agent, the designation of the reference agent can be done either explicitly by a user command or implicitly by an agent detection event.
[0090]
Once the agent is lined up along the path between the two reference agents, other agents can take action. For example, it may be desirable for the agents to converge to a choke point along a path between the two agents. Specifically, if the agent is designed to detect a fire inside one building and deliver the fire suppressant in the best possible way to stop the fire, the effects of the fire suppressant It may be desirable to use the physical structural components of the building to maximize
[0091]
In order to achieve maximum effectiveness, narrow passages are optimal from a fire suppression agent delivery standpoint. Thus, it is possible to determine two reference agents that are present on either side of the narrow passage and to line up the other agent on one path between the reference agents.
[0092]
Next, it would be desirable to have agents lined up along the path search to determine the agent with the least degree of freedom for movement. The agent can be assumed to be in the narrowest part of the passage, and the other agent can be encouraged to head for the agent, and consequently close to the convergence agent Causes local clustering of agents in the cluster. The agent can then deliver the fire suppressant into the narrow passage with maximum effect.
[0093]
When using the lines of a visual communication system, detecting the narrowest part of the path may also be by determining the agent along the path that receives a small number of transmissions from neighboring agents, or Could be achieved by determining the agent receiving the transmission over the smallest angular range.
[0094]
[Further form of the present invention]
Several illustrations illustrating various aspects of the invention are presented to provide a more thorough understanding of how to use them. First, a diagram contrasting the present invention with methods of conventional map-based approaches to ground inference is illustrated in FIG.
[0095]
Next, an illustration of using cost values along a path to determine a path from a departure agent to a target is illustrated in FIG. Next, one example of the use of two distinct signal types to illustrate two different targets is shown in FIG. The alignment of the agents along one path between the two reference agents and the convergence to the choke point are illustrated in FIGS. 7 and 8, respectively.
[0096]
Referring now to FIG. 4, FIG. 4 (a) illustrates some powerful movements through the environment 402 in a search pattern that sends raw or partially processed data back to the central processing resource 404. A conventional means for mapping a space by using a mobile robot or sensor 400 is illustrated. Central processing resource 404 can then edit map 406 and then use the map for other operations.
[0097]
Dynamic changes to the environment will not be picked up until the robot 400 repeats the search. In contrast, as illustrated in FIG. 4 (b), the technique provided by the present invention embodies the map and spreads the map over the entire set of simple agents 100. , Each of which defines its local ground feature. Global properties, including shortest paths, blocked paths, and contingency plans, can be calculated in an error-tolerant and distributed manner, with each member of the plurality of agents 100 contributing to the result.
[0098]
Utilizing the present invention, a distributed group of agents 100 is embedded in the environment to perform both sensing and computational tasks simultaneously. As mentioned above, the agent 100 can be placed in the environment by various means, either randomly or in a specific arrangement. Advantages of using the agent 100 spread throughout an environment include fault tolerance and responsiveness to dynamic changes in the environment.
[0099]
The present invention also provides for the use of a number of relatively small and inexpensive agents that spread throughout an environment, rather than some very expensive robots. Thus, a number of simple agents 100 can be used in one environment with very simple individual computing power to detect dynamic changes there. Since many agents 100 are used, the loss of a single agent 100 has little effect on the overall computing power and can therefore guarantee a high degree of fault tolerance.
[0100]
In addition, communication is only required between the closest neighbors, so there is no need for a high power communication device. Information can be returned to the user 408 by relaying a message, or as the user moves through the environment, the user can obtain data from nearby robots. In addition, the agent 100 can be equipped with means for providing exercise to enable it to actively spread throughout one environment and search for the environment.
[0101]
It is important to note that having multiple small processors spread throughout the environment allows for an improvement in computational speed to the extent that computations can be separated and parallelized. Path planning using existing ground databases takes O (nlogn) time on a single computer, but at worst O (n) time on a network. The best ground point, where the quality of the point can be determined locally, takes O (n) time on a single processor to find it, but O (logn) time on the network. Yes, and in that case most of the time is dedicated to organizing the results. Thus, for large amounts of data, using multiple processors in a distributed network significantly reduces computation time.
[0102]
An illustration of the use of cost value along the path to determine the path from the departure agent to the target is then shown in FIG. With the present invention, it is possible to determine the best path from a starting point 502 inside a building to a target 500. In the illustrated embodiment, the application is a preview inside one building and to locate the target.
[0103]
FIG. 5 (a) is a type of topographic map illustrating a group of agents 100 scattered throughout a building 504, with some members finding the target 500. FIG. It should be noted that the agent 100 has achieved a substantially uniform spacing within the building 504. In this case, the agent 100 includes a suitable sensor for detecting the target.
[0104]
For example, if the target 500 is an invading human, sensors can be provided to detect body temperature, movement, and noise. An illustration depicting a message propagation-like gradient having its cost value encoding the distance to the target is shown in FIG. 5 (b). Upon detecting the target T500, the initiating agent 506, the agent closest to the target T500, sends a signal with a zero cost value of 100. This embodiment illustrates the cost value starting at 100 and decreasing with each hop away from the initiating agent 506.
[0105]
The cost value could be increased very easily from the initiating agent 506 at each hop or, as mentioned earlier, could be varied by a factor representing the local sensor measurement. This could be the distance to the sending agent and could be estimated from the signal strength of the received message.
[0106]
In this embodiment, it is desirable to use a visual communication line between agents, since the internal structure, such as a wall or door, is incorporated into the structural display. Adjacent agents within the visual line of the starting agent 506 receive the signal from the starting agent 506, reduce the cost value, and generate new signals incorporating the reduced cost value. Send the new signal as if it were the initiator.
[0107]
An equal-strength cost value line along which the agent shows all the same cost value, starting at 100 and from 92 to the agent closest to the starting point 502 from the starting agent 506 This is indicated by the line decreasing. This illustrates how the overall cost value of multiple agents forms a gradient-like map. In this case, the line shows the slope of the signal propagating away from the target, while the cost value propagation is the same as the distance information propagation in Dijkstra's shortest path algorithm. The visual communication line is uniquely linked to traversability within the building, and the cost value reflects a similar traverse distance to the target. All agents on the isointensity line are approximately equidistant from the detected target T500. By following the agent whose cost value gradient rises in the previous term, we can expect to find the most efficient path to the target.
[0108]
It may be desirable for the agent 100 to be provided with means for detecting the direction in which the received message was emitted. Whenever a new message is received, the cost value is updated and the corresponding direction to the message source is saved. Using this information, it is possible for each agent to indicate the local direction in which the best is towards the target.
[0109]
A diagram illustrating an agent transmitting a direction signal to an originator based on a cost value message and creating a directed path to the target is shown in FIG. 5 (c). Starting at the starting point 502, the information is provided to the user leading to a single path leading the user to the target T500, while incorporating information about the target location and a passage in the building 504.
[0110]
One example of the use of two distinct signal types to detect two different targets is illustrated in FIG. Although two distinct signal types are used in FIG. 6, any number of signal types can be used as long as each signal type forms a distinct cost value gradient. is there. The cost value gradient includes the distance and direction between each agent and the target or targets. FIG. 6A illustrates that using two message types allows each agent 100 to determine which target is closest. Each of the agents 600, 602, 604, 608 detects which of the first target 610 or the second target 612 is closest.
[0111]
This allows the user to determine which target is closest. In this case, the agent 600 detects the first target 610. The resulting distance to the first target 610 is calculated by each of the other agents 602, 604, 606, 608. At the same time, the agent 608 detects the second target 612. The resulting distance to the second target 612 is also calculated by each of the other agents 602, 604, 606, 608.
[0112]
Distinct message types can be used for each target. For hops, the distances are shown for each agent, separated by a slash, the first distance indicates the distance from the first target 610, and the second distance is the distance from the second target 612. Is shown. Using a distinct message type for each target, the cost value at each agent can be easily compared to determine which target is closest to that agent.
[0113]
Alternatively, a diagram is shown in FIG. 6 (b), where each agent is shown using one message type indicating the distance to the nearest target. In this case, the same message type can be used for both targets, and each agent will only store the distance to the closest agent. As shown in FIG. 6 (b), use a message scheme holding directional information, as discussed above, so that the direction to the nearest target can also be found. Doing it will help.
[0114]
The alignment of the agents along the path between the two reference agents and the convergence to the choke point are illustrated in FIGS. 7 and 8, respectively. As shown in FIG. 7, departure agent 700 and target agent 702 have been designated as reference agents. Each of the reference agents emits a distinct message type to other agents. The other agents 704 generate path vectors for each of the reference agents.
[0115]
The path vector is then the motion vector for each of the other agents 704, along which the line vector is to be lined up on one path 706 between the departure agent 700 and the target agent 702. The other agents 704 can move, but are used to create Note that the arrow represents the path vector between the agent 704 and the reference agent.
[0116]
As the agent 704 moves toward the path 706, the agent 704 arrives at the path and the path 706 moves in a direction opposite to the directions of the departure agent 700 and the target agent 702. The path vector is updated until the vectors are aligned in parallel to.
[0117]
The illustration presented in FIG. 8 illustrates the selection of reference agents to prepare agents to converge along a choke point, and the selection of other agents along one path between them. Utilize aligning. The convergence along the choke point can be performed without first aligning the agents along one path between the reference agents, but by aligning along the path along the path. It supplies the agent at a very high density, allowing more precise detection of the characteristics along the path for the selection of an appropriate choke point.
[0118]
As shown in FIG. 8 (a), the agent at the starting point 800 and the agent at the target point 802 are designated as reference agents, while the path 804 between them is indicated by a dotted line. . In this case, the agent 100 moves toward a point along the bus closest to them, as indicated by the arrow on the agent 100.
[0119]
In FIG. 8B, almost all of the agents are aligned along the path 804. In this case, blocking in the narrowest area between the starting point and the target point is considered desirable because the blocking area is closest to the target point.
[0120]
Thus, two choke points 806 and 808 are specified, and the agent converges on each. It should be noted that the criteria used to determine the choke point in this case are (1) the narrowness of the corridor (area) and (2) the proximity of the corridor to block against the target point. Is a point.
[0121]
Thus, the agent first converges on choke point 806 until no more agents fit in the area and then converges on choke point 808. Another narrow path area 810 is not along the path between the starting point 800 and the target point 802, where the agent does not converge. The area for convergence can be selected in many different ways based on many different criteria. For example, narrowness can be a criterion for blocking an area, while steep slopes can be a criterion against blocking an area. Thus, once an area is sufficiently sloping, the fact that it is narrow is not a problem, and may not need to be blocked.
[0122]
All logic associated with the incorporation of the present invention is typically embodied in the form of software stored in the agent's memory and operated by a processor. However, the logic of the present invention may be fixed in hardware rather than using software. To provide a more complete description of the software and a preferred embodiment thereof, a compact disc is provided herein as an appendix, and some of the features described herein are provided. Or software that performs this.
[Brief description of the drawings]
These and other features and aspects and advantages of the present invention will become better understood with regard to the following description, appended claims, and accompanying drawings where: FIG. That is,
FIG. 1 illustrates one embodiment of a typical agent used in connection with the present invention;
FIG. 2 is a flow diagram illustrating the steps in the method of the present invention, illustrating a method for establishing a best cost value path for at least one initiating agent;
FIG. 3 is a flow diagram illustrating additional steps of the method used to align agents along one path between two reference agents;
FIG. 4 (a) is an illustration illustrating a method of a conventional map-based approach to ground inference;
FIG. 4 (b) is an illustration illustrating the method of the ground inference approach of the present invention, in contrast to the conventional approach illustrated in FIG. 4 (a);
FIG. 5 (a) is a terrain map showing a group of agents scattered within a building with several members searching for their targets;
Figure 5 (b) is an illustration depicting the message being encoded as a gradient with its cost value, encoding the distance to the target;
FIG. 5 (c) illustrates an agent communicating directions to the target based on a cost value message and creating a path direction to the target;
FIG. 6 (a) shows that each agent uses two message types that allow it to determine which target is closest;
FIG. 6 (b) shows that each agent uses one message type to indicate the direction to the nearest target;
FIG. 7 is an illustration showing the use of a path vector to align agents along one path between two reference agents;
FIG. 8 (a) illustrates the creation of one path between two reference agents and the preparation of the other agents to converge to a choke point along the path, FIG. 5 is an explanatory diagram showing that a motion vector is generated from a path vector that prompts another agent to be executed;
FIG. 8B is an explanatory diagram showing a path between reference agents once the agents are arranged along the path;
FIG. 8 (c) is an illustration of the agent after the agent has converged to a choke point along the path;
[Explanation of symbols]
100: agent, 102: communication capability, 104: one set of trucks
106: 1 set of wheels, 108: metal detector, 110: GPS receiver
112: infrared detector, 114: gas chromatography,
400: movable robot or sensor, 402: environment,
404: central processing resource, 406: map, 408: user
500: target, 502: starting point, 504: building,
506: startup agent,
600, 602, 604, 606, 608: Agent,
610: first target, 612: second target,
700: departure agent, 702: departure agent,
704: another agent, 706: path,
800: starting point, 902: target point, 804: pass,

Claims (36)

A method for calculating a characteristic of the physical environment using a plurality of agents located in a physical environment, each agent comprising: a communication capability; a processor coupled to the communication capability; A method comprising utilizing a physical property of inter-agent communication to form a distributed representation of a non-local property of the physical environment, comprising: a memory coupled to a processor;
a. Triggering at least one initiator agent;
b. Sending a signal from the initiating agent to an agent adjacent to the initiating agent;
c. Processing the signal at each of the neighboring agents by a cost value based on local characteristics along a respective path between the initiating agent and each neighboring agent;
d. Selecting a new signal to be transmitted from the received signals;
e. Treating the neighboring agent as a triggering agent of the processed signal and repeating the transmitting and processing steps; and f. At each agent locally retaining at least one accumulated cost value representing information about non-local properties of the physical environment based on the received signal;
The method comprising: 2. The method of calculating characteristics of a physical environment according to claim 1, wherein, in the processing step, the cost value is a hop count, wherein an accumulated hop count from the initiating agent is generated at each agent. 2. The method of claim 1, wherein in the processing step, the hop count is increased at each agent. 2. A method for calculating a characteristic of a physical environment as claimed in claim 1, comprising using the accumulated cost value in a determining step for determining the new signal to be transmitted from the received signal. Item 2. The method according to Item 1. 4. The method of calculating a property of a physical environment as set forth in claim 3, wherein a signal is periodically propagated from the initiating agent across the plurality of agents to determine the cost value for processing the signal. 4. The method of claim 3, including comparing the cost value retained from a past signal with the cost value from a current signal to do so. 4. The method of calculating characteristics of a physical environment as recited in claim 3, wherein the communication path between the agents is a visual path line, thereby transmitting signals across the visual path line. 4. The method of claim 3, wherein the method is used to provide a representation of a traversing presence along the path. 4. A method for calculating a characteristic of a physical environment as set forth in claim 3, wherein the agent includes means for storing the direction in which a particular signal was received over a communication path. 4. The method of claim 3, further comprising: comparing cost values for different paths between two agents; and storing a direction of the path with a best cost. 5. The method of calculating a property of a physical environment according to claim 4, wherein the step of processing the signal further comprises:
a. Sub-step of clocking the cost value when the cost value is held in memory;
b. Specifying a most recently retained time-recorded cost value to be used in processing the signal; and c. Removing the time-recorded cost value from memory after a predetermined time has elapsed;
5. The method according to claim 4, comprising the additional substep of: 8. A method for calculating a characteristic of a physical environment as claimed in claim 7, wherein a plurality of cost values are recorded and ranked from best to worst over time in retaining the cost values in memory. The time value of the time value of the cost value designated for use in processing the signal has expired; and the cost value of the cost value has been specified for use in processing the signal. 8. The method of claim 7, wherein the value is removed from memory, and the next best cost value is designated for use in processing the signal. The method of calculating a property of a physical environment as recited in claim 1, wherein the method further comprises an agent alignment step, wherein the agent alignment step includes:
a. Sub-step of selecting two agents from the plurality of agents to act as reference agents;
b. Generating a cost gradient comprising a vector representation of an optimal cost path between each agent in the plurality of agents and each reference agent;
c. Summing the vector at each agent to create a motion vector for the agent; and d. Moving the agent along the motion vector to align the agent along a path between the reference agents;
The method of claim 1, comprising: 10. The method of calculating a property of a physical environment as recited in claim 9, wherein the agent alignment step includes triggering upon occurrence of a triggering event. 10. The method of calculating a property of a physical environment according to claim 9, wherein the agent alignment step comprises:
a. Selecting a convergence agent; and b. Sub-steps of prompting other agents towards the convergence agent, resulting in local clustering of agents near the convergence agent;
10. The method of claim 9, comprising: 2. The method of calculating a property of a physical environment according to claim 1, wherein the agent further comprises at least one sensor connected to the processor, wherein the sensor provides an output and is used in the signal processing step. The method of claim 1, wherein the performed cost value is developed using the output of the sensor. 13. The method of calculating a property of a physical environment as recited in claim 12, wherein the cost value is used in the determining step to determine a new signal to transmit from the received signal. Item 13. The method according to Item 12. 14. The method of calculating a property of a physical environment as recited in claim 13, wherein a signal including a current signal and at least one past signal is periodically propagated from the initiating agent across the plurality of agents. 14. The method of claim 13, further comprising comparing a cost value retained from a past signal with a cost value from the current signal to determine a cost value for processing the signal. Said method. 14. A method for calculating a property of a physical environment as set forth in claim 13, wherein the agent includes means for storing the direction in which a particular signal was received over a communication path. 14. The method of claim 13, further comprising: comparing cost values for different paths between the two agents; and storing a direction of the path with the best cost. The method of calculating a property of a physical environment as recited in claim 14, wherein the step of processing the signal further comprises:
a. Sub-step of clocking the cost value when the cost value is held in memory;
b. Specifying a most recently retained time-recorded cost value to be used in processing the signal; and c. Removing the time-recorded cost value from memory after a predetermined time has elapsed;
15. The method of claim 14, comprising the additional substep of 17. The method of calculating a characteristic of a physical environment as recited in claim 16, wherein a plurality of cost values are recorded over time when retaining the cost values in memory; and wherein the time-recorded cost values are removed from memory. 17. The method of claim 16 wherein, when done, another historical cost value is selected to adjust the accumulated cost value message. 14. The method of calculating a characteristic of a physical environment as set forth in claim 13, wherein the communication path between the agents is a visual path line, whereby signals over the line of the visual path are transmitted. 14. The method of claim 13, wherein transmitting comprises using to provide a representation of a traversing presence along the path. The method of calculating a property of a physical environment as recited in claim 12, wherein the method further comprises an agent alignment step, wherein the agent alignment step includes:
a. Sub-step of selecting two agents from the plurality of agents to act as reference agents;
b. Generating a cost gradient comprising a vector representation of an optimal cost path between each agent in the plurality of agents and each reference agent;
c. Summing the vector at each agent to create a motion vector for the agent; and d. Moving the agent along the motion vector to align the agent along a path between the reference agents;
13. The method according to claim 12, comprising: 20. The method of calculating characteristics of a physical environment as recited in claim 19, wherein the agent aligning step includes triggering upon the occurrence of a triggering event. 21. The method of calculating a property of a physical environment according to claim 20, wherein the agent alignment step comprises:
a. Selecting a convergence agent; and b. Sub-steps of prompting other agents towards the convergence agent, resulting in local clustering of agents near the convergence agent;
21. The method of claim 20, comprising: An agent used among a plurality of agents located in the physical environment to calculate properties of the physical environment:
a. Communication capability for communicating with other locally spaced agents from among the plurality of agents;
b. Generating a local cost value from the physical characteristics along the path between the agent and another locally spaced agent; and generating the other local value to generate an accumulated cost value. Combining each non-local cost value and the local cost value received in communication with each of the regularly spaced agents to determine the best of the accumulated cost values; A processor coupled with communication capabilities to generate new signals incorporating the best accumulated cost value for transmission to locally spaced agents;
c. A memory connected to the processor for retaining a best accumulated cost value, whereby the plurality of agents are connected to each other to form a distributed representation of a non-local characteristic of the physical environment. Utilizing the physics of communication, at least one of the agents in the plurality of agents is designated as a triggering agent, and the signal is such that the signal is propagated across the plurality of agents and moves away from the triggering agent The agent that has been updated to incorporate the best accumulated cost value, whereby the best path to the initiating agent is saved at each agent in the plurality of agents. 23. An agent used between a plurality of agents located in the physical environment to calculate a property of the physical environment according to claim 22, wherein the non-local cost value is a hop from an initiating agent. A count, wherein the local cost value is an increment of the hop count, and the accumulated cost value is a sum of the local cost value and the best non-local cost value received at the agent to produce an incremental hop count. 23. The agent of claim 22, comprising summing 23. An agent used among a plurality of agents located in the physical environment to calculate a property of the physical environment according to claim 22, wherein the accumulated cost value is calculated from the received signal. 23. The agent of claim 22, including using the processor to determine the best accumulated cost value to send to other locally spaced agents. 25. An agent used between a plurality of agents located in the physical environment to calculate characteristics of the physical environment as set forth in claim 24, wherein the signal is from an initiating agent to the plurality of agents. To determine the best accumulated cost value to be propagated periodically and stored in the memory of the agent to be transmitted to other locally spaced agents. 25. The agent of claim 24, comprising comparing to a stored cost value generated from a newly received signal. 25. An agent used between a plurality of agents located in the physical environment to calculate properties of the physical environment as set forth in claim 24, wherein the communication bus between the agents is a visual agent. Transmission of a signal across the line of the visual path to generate a traversable accumulated cost value indication along the path from the initiating agent to the agent. 25. The agent of claim 24, including using in a processor. 25. An agent used between a plurality of agents located in the physical environment to calculate properties of the physical environment as set forth in claim 24, wherein the agent is further connected to the memory. Means for determining the direction in which a particular signal is being received throughout the communication path, wherein the direction in which the signal having the best cost value was received is stored in the memory. 25. The agent of claim 24, comprising: 26. An agent used among a plurality of agents located within the physical environment to calculate properties of the physical environment as set forth in claim 25, wherein the processor further comprises: Record the time of the accumulated cost value when retained, specify the accumulated cost value recorded the most recently retained time used in processing the signal, and after a predetermined time has elapsed 26. The agent of claim 25, operative to remove accumulated time value recording the time from memory. 29. An agent used between a plurality of agents located in said physical environment to calculate properties of the physical environment as set forth in claim 28, wherein the plurality of accumulated cost values over time. The processor generates another accumulated cost value from memory to be selected to adjust the accumulated cost value message when removing the accumulated cost value recorded in memory and the time recorded from memory. 29. The agent of claim 28, comprising causing the agent to: 23. An agent used between a plurality of agents located in the physical environment to calculate properties of the physical environment according to claim 22, further comprising: an agent connected to the processor. 23. The system of claim 22, wherein at least one sensor is included and the sensor provides an output, and wherein the local cost value is developed by the processor using the output of the sensor. Said agent. 31. An agent used among a plurality of agents located in the physical environment to calculate properties of the physical environment according to claim 30, wherein the accumulated cost value is the received signal. 31. The agent of claim 30, including using in the processor to determine the best accumulated cost value to send to other locally spaced agents from . 32. An agent used between a plurality of agents located in the physical environment to calculate properties of the physical environment as set forth in claim 31, wherein the signal is from a start agent to the plurality of agents. The accumulated cost value stored in the memory of the agent for transmission to other locally spaced agents. 32. The agent of claim 31, comprising comparing to a stored cost value generated from a newly received signal to determine a cost value. 32. An agent used between a plurality of agents located in the physical environment to calculate properties of the physical environment as set forth in claim 31, wherein the communication bus between the agents is visual. Transmitting a signal across the line of the visual path to generate a traversable accumulated cost value indication along one path from the initiating agent to the agent. 32. The agent of claim 31, including being used in the processor of the agent. 32. An agent used between a plurality of agents located in the physical environment to calculate properties of the physical environment as set forth in claim 31, wherein the agent is further connected to the memory. Means for determining the direction in which a particular signal is being received throughout the communication bus, wherein the direction in which the signal having the best cost value was received is stored in the memory. 32. The agent of claim 31, comprising: 33. An agent used between a plurality of agents located in the physical environment to calculate properties of the physical environment as set forth in claim 32, wherein the processor further stores the cost value in memory. Record the time of the accumulated cost value when retained, specify the accumulated cost value recorded the most recently retained time used in processing the signal, and after a predetermined time has elapsed 33. The agent of claim 32, operable to remove from the memory the accumulated cost value that recorded the time. An agent used between a plurality of agents located in said physical environment to calculate properties of the physical environment as set forth in claim 35, wherein the plurality of accumulated cost values over time. The processor generates another accumulated cost value from memory to be selected to adjust the accumulated cost value message when removing the accumulated cost value recorded in memory and the time recorded from memory. 36. The agent of claim 35, comprising causing the agent to:

Images