Multi-agent collaborative photoelectric tracking system based on contract network algorithm and dynamic scheduling method thereof

Abstract

The invention relates to a multi-agent collaborative photoelectric tracking system based on a contract net algorithm and a dynamic scheduling method thereof, and aims to solve the technical problems of low matching efficiency, low observation precision and low system robustness of a multi-agent photoelectric tracking system in the prior art. According to the multi-agent cooperative photoelectric tracking system based on the contract network algorithm and the dynamic scheduling method thereof, the central node unit is in bidirectional communication connection with the agent unit and is used for sending the task execution request to the agent unit after monitoring that the target enters the preset measurement area; after receiving a task execution request, the agent units calculate cost data and transmit the cost data to the central node unit, the central node unit carries out bid invitation on the agent units according to the cost data, and the agent units winning bids carry out dynamic allocation on monitoring targets.

Description

Technical Field

[0001] The present invention relates to a photoelectric monitoring network joint measurement system, and in particular to a multi-agent cooperative photoelectric tracking system based on a contract network algorithm and a dynamic scheduling method thereof. Background Art

[0002] The multi-agent system used in optoelectronic tracking usually adopts active positioning system or passive positioning system. The active positioning system is easy to expose its own position when tracking the target, and is thus vulnerable to countermeasures. Therefore, the passive positioning system is often used in the joint measurement of the ground-based optoelectronic monitoring network. The passive positioning system usually uses multiple optoelectronic theodolites to achieve positioning. It can not only independently complete the detection of various measurement targets, but also be combined with the radar detection system in complex environments to improve the measurement capability.

[0003] The existing multi-agent photoelectric tracking passive positioning system combines multiple photoelectric theodolites and integrates peripheral modules to intelligently upgrade the individual devices. This system forms a collaborative observation multi-agent photoelectric tracking system, which can enhance the observation capability of the covered measurement space and achieve stable, accurate, and fast effective tracking. However, the existing multi-agent photoelectric tracking system has the following problems when performing task scheduling:

[0004] 1. The task scheduling method uses a fixed device-target pairing method. Although this method ensures target tracking accuracy, it is difficult to adapt to scenarios with multiple targets moving dynamically over a large area at high speed. Therefore, it suffers from low matching efficiency in such scenarios.

[0005] 2. The many factors affecting dynamic scheduling are not fully considered during task scheduling, resulting in fluctuations in observation accuracy and thus low observation accuracy;

[0006] 3. During the task scheduling process, due to communication failures between individual nodes and the central node, and the difficulty in adjusting task conflicts, the central node is overloaded with tasks, resulting in low system robustness. Summary of the Invention

[0007] The purpose of the present invention is to solve the technical problems of low matching efficiency, low observation accuracy and low system robustness of the existing multi-agent photoelectric tracking system, and to provide a multi-agent collaborative photoelectric tracking system based on the contract network algorithm and its dynamic scheduling method.

[0008] To achieve the above objectives, the technical solutions provided by the present invention are as follows:

[0009] A multi-agent cooperative photoelectric tracking system based on the contract network algorithm has the following characteristics:

[0010] It includes central node unit and intelligent agent unit;

[0011] The central node unit includes a target monitoring module and a task management module; the intelligent agent unit includes an intelligent agent module and a decision module, and the intelligent agent module includes multiple intelligent agents;

[0012] The sending end of the target monitoring module is connected to the first receiving end of the task management module, and is used to send a task execution request to the task management module after monitoring that the target to be measured enters the preset measurement area;

[0013] The first transmission end of the task management module and the first transmission end of the decision module are bidirectionally connected for communication, and are used to send the cost calculation task to the decision module after receiving the task execution request;

[0014] The second transmission end of the decision module and the first transmission end of the agent module are bidirectionally connected for communication, and are used to send the received cost calculation task to the agent module. After receiving the cost calculation task, the agent module calculates the cost data and transmits the calculated cost data to the decision module. The decision module transmits the cost data to the task management module via the first transmission end.

[0015] The second transmission end of the task management module and the second transmission end of the intelligent agent module are bidirectionally connected for communication, and are used to generate a bid document based on the received cost data and send the bid document to the intelligent agent module, and to bid for the intelligent agent module. The intelligent agent module generates a bid package for bidding based on the bid document and sends the bid package to the task management module. The task management module determines the winning intelligent agent information in the intelligent agent module based on the bid package and sends the winning intelligent agent information to the intelligent agent module. The intelligent agent module dynamically schedules the intelligent agent based on the winning intelligent agent information, thereby completing the allocation of target measurement tasks. The intelligent agent in the intelligent agent module tracks the target to be measured based on the allocation result.

[0016] Furthermore, the intelligent agent module includes an intelligent agent node submodule and an edge computing submodule. The intelligent agent node submodule includes the multiple intelligent agents, and its first transmitting end is connected to the receiving end of the edge computing submodule, and is used to transmit the intelligent agent data in the intelligent agent node submodule to the edge computing submodule; the second transmitting end of the decision module and the first transmitting end of the edge computing submodule are bidirectionally connected to each other, and the first transmitting end of the edge computing submodule constitutes the first transmitting end of the intelligent agent module, and is used to send the cost calculation task to the edge computing submodule. The edge computing submodule calculates the cost data of the intelligent agent based on the received intelligent agent data and transmits the cost data to the decision module;

[0017] The second transmission end of the task management module and the first transmission end of the intelligent agent node submodule are bidirectionally connected for communication. The first transmission end of the intelligent agent node submodule constitutes the second transmission end of the intelligent agent module, which is used to generate a bid document based on the received cost data and send the bid document to the intelligent agent node submodule, and to bid for the intelligent agent node submodule. The intelligent agent node submodule generates a bid package for bidding based on the bid document, and sends the bid package to the task management module. The task management module determines the winning intelligent agent information in the intelligent agent node submodule based on the bid package, and sends the winning intelligent agent information to the intelligent agent node submodule. The intelligent agent node submodule dynamically schedules the intelligent agent based on the winning intelligent agent information, thereby completing the allocation of target measurement tasks. The intelligent agent in the intelligent agent module tracks the target to be measured based on the allocation result.

[0018] Furthermore, the central node unit also includes a status monitoring module, the receiving end of the status monitoring module is connected to the second sending end of the intelligent agent node sub-module, and is used to receive the status information of the intelligent agent sent by the intelligent agent node sub-module, and the sending end of the status monitoring module is connected to the second receiving end of the task management module, and is used to send the received status information of the intelligent agent to the task management module. The task management module determines the winning intelligent agent information based on the status information of the intelligent agent and the bidding package.

[0019] Furthermore, it also includes a time synchronization unit, which is respectively connected to the target monitoring module, task management module, status monitoring module, decision module, intelligent node sub-module and edge computing sub-module for time synchronization.

[0020] At the same time, the present invention also provides a dynamic scheduling method for multi-agent cooperative photoelectric tracking based on the contract net algorithm, which adopts the multi-agent cooperative photoelectric tracking system based on the contract net algorithm. The method is special in that it includes the following steps:

[0021] Step 1: Use the target monitoring module to perform target monitoring. After the target monitoring module detects that the target to be measured enters the preset measurement area, it sends a task execution request to the task management module;

[0022] Step 2: After receiving the task execution request, the task management module sends the cost calculation task to the decision module;

[0023] Step 3: After receiving the cost calculation task request, the decision module sends the cost calculation task request to the agent module;

[0024] Step 4: After receiving the cost calculation task request, the agent module performs cost calculation and transmits the calculated cost data to the decision module, which then transmits the received cost data to the task management module.

[0025] Step 5. The task management module generates a bid document based on the received cost data and sends the bid document to the intelligent agent module to bid for the intelligent agent module. The intelligent agent module generates a bid package for bidding based on the bid document and sends the bid package to the task management module. The task management module determines the winning intelligent agent information in the intelligent agent module based on the bid package and sends the winning intelligent agent information to the intelligent agent module. The intelligent agent module dynamically schedules the intelligent agent based on the winning intelligent agent information, thereby completing the allocation of target measurement tasks. The intelligent agent in the intelligent agent module tracks the target to be measured based on the allocation results.

[0026] Furthermore, step 5 is specifically as follows:

[0027] Step 5.1: The task management module generates a bid based on the current target information detected by the target monitoring module, the agent's observation method, and the received cost data, and sends it to the agent node submodule to bid for the agents in the agent node submodule;

[0028] Step 5.2: The agent node submodule generates a bidding package for bidding based on the received bid document and sends it to the task management module;

[0029] Step 5.3: The agent node submodule obtains the agent's status information and sends it to the status monitoring module. The status monitoring module sends the received agent's status information to the task management module.

[0030] Step 5.4: The task management module selects the agent with the lowest measurement cost for the target to be measured as the winning agent based on the received bid package and agent status information, and sends the winning agent information to the agent node submodule;

[0031] Step 5.5: After the agent node submodule receives the winning agent information, the winning agents will communicate with each other to obtain the tasks assigned to each agent. The agent with multiple tasks will enter a reallocation queue. After retaining the task with the highest priority, the remaining tasks will be reallocated excluding the agent until all tasks have corresponding agents to monitor them. The dynamic scheduling of the agent is completed, thereby completing the allocation of target measurement tasks. The agent in the agent module tracks the target to be measured according to the allocation results.

[0032] Furthermore, step 4 is specifically as follows:

[0033] Step 4.1, the agent node submodule transmits the acquired agent data to the edge computing submodule;

[0034] Step 4.2: After receiving the cost calculation task request, the edge computing submodule performs cost calculation based on the received data of each intelligent agent and transmits the calculated cost data to the decision module;

[0035] Step 4.3: The decision module transmits the received cost data to the task management module.

[0036] Furthermore, the cost data calculation process in step 4.2 is as follows:

[0037] Step 4.21. Calculate the agent's health cost p1

[0038]

[0039] In the formula, α is the weight of the longest remaining working time of the device, β is the load balancing weight, T max,i is the maximum working time of the agent, Tasktime i is the working time of the ith agent, e k is the Sigmoid function, k is the steepness coefficient, is the average load time;

[0040] Step 4.22: Calculate the attenuation cost p2 of the distance between the agent and the target

[0041]

[0042] Where d is the Euclidean distance between the agent and the target observation task, which satisfies: d = ‖x task -x agent ‖,x task is the coordinate of the target, x agent is the coordinate of the agent; d max is the maximum effective observation distance of the intelligent device, p2∈[0,1);

[0043] Step 4.23, calculate the agent intersection measurement cost p3

[0044]

[0045] Where a and b are any two intersecting agents, p a3 , p b3 are their corresponding intersection measurement costs, θ is the angle between the agent and the target, μ is the ideal intersection angle, which satisfies: μ = 90°, η i is the angle measurement accuracy of the device, σ i is the dynamic standard deviation, v is the standard deviation, which satisfies the following formula:

[0046] σ i =30+10(1-ηi ),η i ∈[0,1];

[0047] Step 4.24, calculate the impact cost of the agent's task switching as p4

[0048]

[0049] Where γ is the device switching cost, γ = 0.4, T remaining is the remaining time of the current task, T total is the total task time;

[0050] Step 4.25: Calculate the agent task learning cost p5

[0051]

[0052] Where, is the past matching degree of agent i to the task type k, ω is the learning ability level parameter of the agent, The completion status of the current measurement task;

[0053] Step 4.26: Calculate the total bidding cost p6 of the agent

[0054] p6=(v1·p1+v2·p2+v3·p3+ν4·p4+ν5·p5)·v6

[0055] Where ν1, ν2, ν3, ν4, ν5, and v6 are the weight values ​​corresponding to p1, p2, p3, p4, p5, and p6, respectively. They are all typical values ​​of the measurement test, and the weight constraint is set as: ν1+ν2+ν3+ν4+ν5=1.

[0056] Beneficial effects of the present invention:

[0057] 1. The present invention discloses a multi-agent cooperative photoelectric tracking system based on a contract net algorithm and its dynamic scheduling method. By combining a multi-agent system with an photoelectric tracking system to form a multi-agent cooperative photoelectric tracking system, dynamic task allocation is transformed into a market game mechanism between agents. After task allocation using the contract net algorithm, conflicts are avoided through negotiation between agents. This system achieves real-time dynamic task allocation for multiple targets, effectively reducing the allocation time of measurement tasks and improving the reliability and effectiveness of target monitoring data. This solves the problems of low matching efficiency, low observation accuracy, and low system robustness in existing multi-agent photoelectric tracking systems.

[0058] 2. The present invention proposes a multi-agent cooperative optoelectronic tracking system and its dynamic scheduling method based on a contract net algorithm. By setting agent health status costs, agent-target distance attenuation costs, agent intersection measurement costs, agent task switching impact costs, agent task learning cost, and agent total bid costs, along with a weighted constraint model knowledge base, this system evaluates the agent's target monitoring effectiveness. Furthermore, by organically combining the constraint model with the multi-agent system contract net algorithm, this system implements intelligent networked detection for the optoelectronic tracking system, addressing the issue of low observation accuracy.

[0059] 3. The present invention provides a multi-agent collaborative optoelectronic tracking system based on a contract network algorithm and a dynamic scheduling method thereof. By setting a negotiation mechanism between agents and adopting a secondary negotiation mechanism after task allocation, hierarchical allocation of multiple targets is achieved, the task allocation time of multiple targets is reduced, and timely response is ensured when multiple targets are allocated. At the same time, the negotiation does not go through the central node unit, which reduces the information load of the central node unit during the secondary negotiation and solves the problem of low matching efficiency in the large-scale dynamic high-speed motion scenario of multiple targets. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 This is a system block diagram of an embodiment of a multi-agent cooperative photoelectric tracking system based on a contract network algorithm of the present invention;

[0061] Figure 2 This is a communication topology diagram of an embodiment of a multi-agent cooperative photoelectric tracking system based on a contract network algorithm of the present invention;

[0062] Figure 3 This is a flow chart of an embodiment of a dynamic scheduling method for multi-agent cooperative photoelectric tracking based on a contract network algorithm of the present invention;

[0063] Figure 4 This is a time series diagram of task allocation at t=0s, t=3.0s, t=4.0s, t=5.0s, t=6.0s, and t=7.0s in an embodiment of a dynamic scheduling method for multi-agent collaborative photoelectric tracking based on a contract network algorithm of the present invention. DETAILED DESCRIPTION

[0064] The present invention will be further described below with reference to the accompanying drawings and examples.

[0065] This embodiment is a multi-agent cooperative photoelectric tracking system based on the contract network algorithm. Figure 1 and Figure 2As shown in the figure, it includes a central node unit, an agent unit and a time synchronization unit. The central node unit includes a target monitoring module, a task management module and a status monitoring module; the agent unit includes an agent module and a decision module, and the agent module includes an agent node submodule and an edge computing submodule. In a distributed multi-agent system, an independent central node unit is usually not set up. The central node unit is generally undertaken by an agent in the system, which is convenient for switching according to the situation. However, in the field of joint measurement of ground-based optoelectronic monitoring networks, the optoelectronic theodolite, as an agent unit, undertakes more important precision measurement tasks. In order to reduce the communication pressure and the load of the optoelectronic theodolite equipment, it is necessary to set up an independent central node unit for dynamic scheduling. In addition, after adding necessary peripherals (such as the decision module and the edge computing submodule), the optoelectronic theodolite equipment can be upgraded to an agent unit to enable it to have decision-making capabilities.

[0066] The target monitoring module's transmitting end is connected to the first receiving end of the task management module. When the target monitoring module detects that the target to be measured enters the preset measurement area, it sends a task execution request to the task management module. The first transmitting end of the task management module is bidirectionally connected to the first transmitting end of the decision module. After receiving the task execution request, the task management module sends the cost calculation task to the decision module.

[0067] The second transmission end of the decision module and the first transmission end of the edge computing submodule are bidirectionally connected for communication, and are used to send the cost calculation task to the edge computing submodule. The first transmitting end of the intelligent node submodule and the receiving end of the edge computing submodule are connected to transmit the information of each intelligent agent in the intelligent node submodule to the edge computing submodule. After the edge computing submodule receives the cost calculation task, it calculates the cost data of each intelligent agent based on the received information of each intelligent agent, and transmits the cost data of each intelligent agent to the decision module. The decision module transmits the cost data of each intelligent agent to the task management module through the first transmission end.

[0068] The second transmission end of the task management module is bidirectionally connected to the first transmission end of the intelligent agent node submodule. The task management module generates a bid document based on the received cost data of each intelligent agent, the current target information detected by the radar, and the observation method of the intelligent agent, and sends the bid document to the intelligent agent node submodule to bid for each intelligent agent in the intelligent agent node submodule. The intelligent agent node submodule sends the bid package for bidding generated based on the bid document to the task management module.

[0069] The receiving end of the state monitoring module is connected to the second sending end of the intelligent agent node submodule, and the sending end of the state monitoring module is connected to the second receiving end of the task management module, and is used to receive the state information of the intelligent agent sent by the intelligent agent node submodule and send it to the task management module;

[0070] The task management module determines the winning agent information based on the agent's status information and bidding package, and sends the winning agent information to the agent node submodule. The agent node submodule completes the allocation of target measurement tasks based on the winning agent information.

[0071] At the same time, the transmission end of the time synchronization unit is respectively connected to the first transmission end of the target monitoring module, the third transmission end of the task management module, the first transmission end of the status monitoring module, the third transmission end of the decision module, the second transmission end of the intelligent node sub-module and the second transmission end of the edge computing sub-module, so as to synchronize the time of each module to reduce delay.

[0072] This embodiment provides a dynamic scheduling method for a multi-agent cooperative photoelectric tracking system based on a contract network algorithm. Figure 3 As shown in the figure, assume that in a fixed-size monitoring environment, multiple photoelectric theodolites are deployed on the ground to perform precise measurements on two dynamic targets of different priorities passing through the environment. As time changes, the positions of the dynamic targets also change, and different photoelectric theodolites are required to complete precise measurements at different times. To ensure measurement accuracy, the positions of the photoelectric theodolites must remain unchanged. The location set of the photoelectric theodolites in the mission site is:

[0073] N d ={N1,N2,N3,...,N m}

[0074] Where N d is a finite set. The element N in the set stores the position information of the photoelectric theodolite. m is the number of photoelectric theodolites. The photoelectric theodolite can be upgraded to an intelligent unit after adding necessary peripherals. At this time, N d is a collection of intelligent agents.

[0075] There are two target observation tasks, namely the aircraft and the impactor. The impactor has a higher priority than the aircraft. There are 12 photoelectric theodolites, which are randomly distributed in various positions in space. The preset trajectory of the target in the simulation is consistent with the actual trajectory. The trajectory of the aircraft is a straight line, and the trajectory of the impactor is a parabola. Their trajectories meet the following requirements:

[0076]

[0077] Where trajectory1 is the aircraft trajectory, trajectory2 is the DD trajectory, t is the timing time, and x, y, and z are the trajectory coordinates.

[0078] Based on the above scenario, the calculation steps of the dynamic scheduling method of the multi-agent cooperative photoelectric tracking system embodiment based on the contract network algorithm of the present invention are as follows:

[0079] Step 1: After detecting that the aircraft and the impact object have entered the preset measurement area, the target monitoring module sends a task execution request to the task management module;

[0080] Step 2: After receiving the task execution request, the task management module sends the cost calculation task to the decision module;

[0081] Step 3: After receiving the cost calculation task request, the decision module sends the cost calculation task request to the edge computing submodule. The edge computing submodule calculates the cost data of each agent. In this embodiment, the cost data of the jth agent can be expressed as:

[0082]

[0083] Where, is the health status cost of a single device, is the close priority cost, Measuring the cost for intersection, The cost of task switching, is the learning cost of the task, is the comprehensive total consideration;

[0084] Step 3.1: Calculate the health cost of the jth agent based on the total running time of a single device from the start of the target measurement task to the time it is called to perform the target measurement task.

[0085]

[0086] In the formula, α is the weight of the longest remaining working time of the device, β is the load balancing weight, T max,i is the maximum working time of the agent, Tasktime i is the working time of the ith agent, e k is the Sigmoid function, k is the steepness coefficient, is the average load time;

[0087] Step 3.2: Calculate the short-distance priority cost between the jth agent and the target

[0088]

[0089] Where d is the Euclidean distance between the agent and the target observation task, which satisfies: d = ‖x task -x agent ‖, where x task is the coordinate of the target, x agent is the coordinate of the agent; d maxis the maximum effective observation distance of the intelligent device,

[0090] Step 3.3: Set the two devices as a pair of intelligent agents for collaborative observation. Calculate the measured intersection angle based on the target position information and the intelligent agent position information. Calculate the intersection measurement cost of the jth intelligent agent based on the measured intersection angle between the intelligent agent combination and the target.

[0091]

[0092] Where a is any agent except agent b, θ is the angle between the agent and the target, μ is the ideal intersection angle, which satisfies: μ = 90°, η i is the angle measurement accuracy of the device (1 is the highest accuracy), σ i is the dynamic standard deviation, σ is the standard deviation, which satisfies the following formula:

[0093] σ i =30+10(1-η i ),η i ∈[0,1];

[0094] Step 3.4: Calculate the impact cost of task switching of the jth agent as

[0095]

[0096] Where γ is the device switching cost, γ = 0.4, T remaining is the remaining time of the current task, T total is the total task time;

[0097] Step 3.5: Formulate dynamic learning adaptability rules and calculate the task learning cost of the jth agent

[0098]

[0099] Where, is the past matching degree of agent i to the task type k, ω is the learning ability level parameter of the agent, The completion status of the current measurement task;

[0100] Step 3.6: Calculate the total cost of the jth agent when performing the optical measurement task

[0101]

[0102] In the formula, the weight constraint is set as follows: ν1+ν2+ν3+ν4+ν5=1, where ν1, ν2, ν3, ν4, and ν5 are all typical values ​​from the measurement test. If the actual measured distance exceeds the maximum detection distance of the intelligent agent device, the intelligent agent combination cannot work normally. In this case, the total cost is 0 and no bid is made.

[0103] The goal of the dynamic scheduling problem of a multi-agent optoelectronic tracking system is to reasonably allocate tasks to dynamic targets at different positions at different times so that they can achieve the best precision measurement level at different times. The mathematical model of the dynamic scheduling problem of a multi-agent optoelectronic tracking system is:

[0104]

[0105] Where N i is the number of cost types in the contract net algorithm, p ij is the cost of the corresponding agent, x ij =0 indicates that the task is not executed, x ij =2 indicates execution of the task, L i is the number of agents performing the task, and st is the constraint condition;

[0106] Step 4: The edge computing submodule transmits the calculated cost data to the decision module, and the decision module transmits the received cost data to the task management module;

[0107] Step 5: The task management module generates a bid based on the current target information detected by the radar, the agent's observation method, and the received cost data;

[0108] Step 6: The agent node submodule obtains the status information of each agent and sends it to the status monitoring module. The status monitoring module sends the received status information of each agent to the task management module. Since the photoelectric theodolite has an observation range, if the target exceeds the observation range, it cannot be observed and the observation cost is 0. In this case, it will interfere with the global optimal solution. Therefore, the agents that cannot work are first removed according to the observation situation. Therefore, the status information of each agent includes both the agents that work normally and the agents that do not work normally.

[0109] Step 7: The task management module sends the bid document obtained in step 5 and the status information, target location, and task requirements of each agent obtained in step 6 to the agent node submodule. The agent in the agent node submodule conducts bidding. Each agent is effectively paired and combined based on its status information, target location, and task requirements. After pairing with each other, the paired agents generate a bid package for bidding according to the price format in the bid document and send it to the task management module.

[0110] Step 8: The task management module analyzes the measurement costs based on the received bid package, compares the costs of each agent based on the measurement costs, selects the agent with the lowest measurement cost for each target as the winning agent, and sends the winning agent information to the agent node submodule;

[0111] Step 9. After the agent node submodule receives the winning agent information, the winning agents communicate with each other to obtain the tasks assigned to each agent. The agent with multiple tasks will enter a reallocation queue. After retaining the highest priority target, the remaining tasks will exclude the agent for reallocation until each agent has the corresponding highest priority target, and the dynamic scheduling of the agent is completed.

[0112] Figure 4 The figure shows the matching of different target and equipment combinations at different times (t=0s, t=3.0s, t=4.0s, t=5.0s, t=6.0s, t=7.0s). In the figure, gray dots represent unselected equipment, blue dots represent aircraft observation stations, red dots represent impactor observation stations, blue five-pointed stars represent aircraft, and red five-pointed stars represent impactors. When aircraft and impactors appear, the algorithm can assign the optimal combination of intelligent agents to the targets. Experimental results show that the contract network task allocation algorithm can realize intelligent planning, deployment and measurement of multiple targets and temporary targets, and the measurement results can be further quantified by cost, so as to maximize the sum of the total cost of all equipment in the task redistribution process, thereby finding the optimal task allocation and scheduling solution.

[0113] Any content not described in detail in this specification belongs to the prior art known to those skilled in the art. The above examples are provided for illustrative purposes only and are not intended to limit the scope of the present invention. The scope of the present invention is defined by the appended claims. All equivalent substitutions and modifications that do not depart from the spirit and principles of the present invention are intended to be within the scope of the present invention.

Claims

1. A multi-agent cooperative photoelectric tracking system based on a contract net algorithm, characterized by: It includes central node unit and intelligent agent unit; The central node unit includes a target monitoring module and a task management module; the intelligent agent unit includes an intelligent agent module and a decision module, and the intelligent agent module includes multiple intelligent agents; The sending end of the target monitoring module is connected to the first receiving end of the task management module, and is used to send a task execution request to the task management module after monitoring that the target to be measured enters the preset measurement area; The first transmission end of the task management module and the first transmission end of the decision module are bidirectionally connected for communication, and are used to send the cost calculation task to the decision module after receiving the task execution request; The second transmission end of the decision module and the first transmission end of the agent module are bidirectionally connected for communication, and are used to send the received cost calculation task to the agent module. After receiving the cost calculation task, the agent module calculates the cost data and transmits the calculated cost data to the decision module. The decision module transmits the cost data to the task management module via the first transmission end. The second transmission end of the task management module and the second transmission end of the intelligent agent module are bidirectionally connected for communication, and are used to generate a bid document based on the received cost data and send the bid document to the intelligent agent module, and to bid for the intelligent agent module. The intelligent agent module generates a bid package for bidding based on the bid document and sends the bid package to the task management module. The task management module determines the winning intelligent agent information in the intelligent agent module based on the bid package and sends the winning intelligent agent information to the intelligent agent module. The intelligent agent module dynamically schedules the intelligent agent based on the winning intelligent agent information, thereby completing the allocation of target measurement tasks. The intelligent agent in the intelligent agent module tracks the target to be measured based on the allocation result.

2. The multi-agent cooperative photoelectric tracking system based on the contract net algorithm according to claim 1 is characterized in that: The intelligent agent module includes an intelligent agent node submodule and an edge computing submodule. The intelligent agent node submodule includes the multiple intelligent agents. Its first transmitting end is connected to the receiving end of the edge computing submodule, and is used to transmit the intelligent agent data in the intelligent agent node submodule to the edge computing submodule; the second transmitting end of the decision module and the first transmitting end of the edge computing submodule are bidirectionally connected for communication. The first transmitting end of the edge computing submodule constitutes the first transmitting end of the intelligent agent module, and is used to send the cost calculation task to the edge computing submodule. The edge computing submodule calculates the cost data of the intelligent agent based on the received intelligent agent data and transmits the cost data to the decision module; The second transmission end of the task management module and the first transmission end of the intelligent agent node submodule are bidirectionally connected for communication. The first transmission end of the intelligent agent node submodule constitutes the second transmission end of the intelligent agent module, which is used to generate a bid document based on the received cost data and send the bid document to the intelligent agent node submodule, and to bid for the intelligent agent node submodule. The intelligent agent node submodule generates a bid package for bidding based on the bid document, and sends the bid package to the task management module. The task management module determines the winning intelligent agent information in the intelligent agent node submodule based on the bid package, and sends the winning intelligent agent information to the intelligent agent node submodule. The intelligent agent node submodule dynamically schedules the intelligent agent based on the winning intelligent agent information, thereby completing the allocation of target measurement tasks. The intelligent agent in the intelligent agent module tracks the target to be measured based on the allocation result.

3. The multi-agent cooperative photoelectric tracking system based on the contract net algorithm according to claim 2 is characterized in that: The central node unit also includes a status monitoring module, the receiving end of the status monitoring module is connected to the second sending end of the intelligent agent node sub-module, and is used to receive the status information of the intelligent agent sent by the intelligent agent node sub-module. The sending end of the status monitoring module is connected to the second receiving end of the task management module, and is used to send the received status information of the intelligent agent to the task management module. The task management module determines the winning intelligent agent information based on the status information of the intelligent agent and the bidding package.

4. The multi-agent cooperative photoelectric tracking system based on the contract net algorithm according to claim 3 is characterized by: It also includes a time synchronization unit, which is respectively connected to the target monitoring module, task management module, status monitoring module, decision module, intelligent node sub-module and edge computing sub-module for time synchronization.

5. A dynamic scheduling method for multi-agent cooperative photoelectric tracking based on a contract net algorithm, using the multi-agent cooperative photoelectric tracking system based on a contract net algorithm according to any one of claims 1 to 4; characterized in that: The following steps are involved: Step 1: Use the target monitoring module to perform target monitoring. After the target monitoring module detects that the target to be measured enters the preset measurement area, it sends a task execution request to the task management module; Step 2: After receiving the task execution request, the task management module sends the cost calculation task to the decision module; Step 3: After receiving the cost calculation task request, the decision module sends the cost calculation task request to the agent module; Step 4: After receiving the cost calculation task request, the agent module performs cost calculation and transmits the calculated cost data to the decision module, which then transmits the received cost data to the task management module. Step 5. The task management module generates a bid document based on the received cost data and sends the bid document to the intelligent agent module to bid for the intelligent agent module. The intelligent agent module generates a bid package for bidding based on the bid document and sends the bid package to the task management module. The task management module determines the winning intelligent agent information in the intelligent agent module based on the bid package and sends the winning intelligent agent information to the intelligent agent module. The intelligent agent module dynamically schedules the intelligent agent based on the winning intelligent agent information, thereby completing the allocation of target measurement tasks. The intelligent agent in the intelligent agent module tracks the target to be measured based on the allocation results.

6. The dynamic scheduling method for multi-agent cooperative photoelectric tracking based on the contract net algorithm according to claim 5 is characterized in that: Step 5 is as follows: Step 5.1: The task management module generates a bid based on the current target information detected by the target monitoring module, the agent's observation method, and the received cost data, and sends it to the agent node submodule to bid for the agents in the agent node submodule; Step 5.2: The agent node submodule generates a bidding package for bidding based on the received bid document and sends it to the task management module; Step 5.3: The agent node submodule obtains the agent's status information and sends it to the status monitoring module. The status monitoring module sends the received agent's status information to the task management module. Step 5.4: The task management module selects the agent with the lowest measurement cost for the target to be measured as the winning agent based on the received bid package and agent status information, and sends the winning agent information to the agent node submodule; Step 5.5: After the agent node submodule receives the winning agent information, the winning agents will communicate with each other to obtain the tasks assigned to each agent. The agent with multiple tasks will enter a reallocation queue. After retaining the task with the highest priority, the remaining tasks will be reallocated excluding the agent until all tasks have corresponding agents to monitor them. The dynamic scheduling of the agent is completed, thereby completing the allocation of target measurement tasks. The agent in the agent module tracks the target to be measured according to the allocation results.

7. The dynamic scheduling method for multi-agent cooperative photoelectric tracking based on the contract net algorithm according to claim 6 is characterized in that: Step 4 is as follows: Step 4.1, the agent node submodule transmits the acquired agent data to the edge computing submodule; Step 4.2: After receiving the cost calculation task request, the edge computing submodule performs cost calculation based on the received data of each intelligent agent and transmits the calculated cost data to the decision module; Step 4.3: The decision module transmits the received cost data to the task management module.

8. The dynamic scheduling method for multi-agent cooperative photoelectric tracking based on the contract net algorithm according to claim 7 is characterized in that: The calculation process of the cost data in step 4.2 is as follows: Step 4.

21. Calculate the agent's health cost p1 In the formula, α is the weight of the longest remaining working time of the device, β is the load balancing weight, T max,i is the maximum working time of the agent, Tasktime i is the working time of the ith agent, e k is the Sigmoid function, k is the steepness coefficient, is the average load time; Step 4.22: Calculate the attenuation cost p2 of the distance between the agent and the target Where d is the Euclidean distance between the agent and the target observation task, which satisfies: d = ‖x task -x agent ‖, where x task is the coordinate of the target, x agent is the coordinate of the agent; d max is the maximum effective observation distance of the intelligent device, p2∈[0,1); Step 4.23, calculate the agent intersection measurement cost p3 Where a and b are any two intersecting agents, p a3 , p b3 are their corresponding intersection measurement costs, θ is the angle between the agent and the target, μ is the ideal intersection angle, which satisfies: μ = 90°, η i is the angle measurement accuracy of the device, σ i is the dynamic standard deviation, σ is the standard deviation, which satisfies the following formula: s i =30+10(1-th i ),or i ∈[0,1]; Step 4.24, calculate the impact cost of the agent's task switching as p4 Where γ is the device switching cost, γ = 0.4, T remaining is the remaining time of the current task, T total is the total task time; Step 4.25: Calculate the agent task learning cost p5 Where, is the past matching degree of agent i to the task type k, ω is the learning ability level parameter of the agent, The completion status of the current measurement task; Step 4.26: Calculate the total bidding cost p6 of the agent p6=(v1·p1+v2·p2+ν3·p3+v4·p4+v5·p5)·v6 Where ν1, ν2, ν3, ν4, ν5, and v6 are the weight values ​​corresponding to p1, p2, p3, p4, p5, and p6, respectively. They are all typical values ​​of the measurement test, and the weight constraint is set as: ν1+ν2+ν3+ν4+ν5=1.

Images