Dual task inspection method for road continuous monitoring and handling of incidents

Abstract

The application provides a double-task inspection method for road continuous monitoring and handling of emergencies. The method solves the problem that a group of robots cannot monitor the environment and handle emergencies at the same time in a road network, and can be applied to forest fire, police patrol, environmental governance and other inspection scenes. In order to ensure that the emergency in the application scene can be handled in priority, the application innovatively constructs a double-task priority model, and the group of robots can adaptively adjust the income according to the monitored emergency, and quickly respond to handle the emergency.

Description

Technical Field

[0001] This invention relates to the field of collaborative inspection by swarm robots, and in particular to a dual-task inspection method for continuous road monitoring and handling of emergencies. Background Technology

[0002] Collaborative inspection by swarm robots refers to a group of robots patrolling roads or open areas to obtain situational information and ensure environmental safety. It is widely used in epidemic prevention, area defense, disaster management, and environmental monitoring. Existing collaborative inspection methods by swarm robots are divided into two categories: continuous inspection and adversarial inspection.

[0003] Continuous inspection requires robots to visit key nodes in the environment as frequently as possible and is applied in crowd movement monitoring, agricultural management, and data acquisition. The goal of continuous inspection is to monitor for lingering events by reducing the time interval between robot visits to each node. Therefore, the primary evaluation criteria for this type of inspection are time and movement cost. Existing continuous inspection methods share a common characteristic: robots use deterministic strategies to continuously and cyclically visit viewpoints. They assume that environmental conditions change regularly over time. Therefore, they operate in a stable environment and do not consider unforeseen events.

[0004] Adversarial inspection focuses on preventing intruders from entering, polluting, and damaging the environment, and is widely used in territorial defense and border control. In these highly adversarial scenarios, intruders can determine the timing and location of their attacks by analyzing the strategies of defensive robots. Therefore, adversarial inspection methods employ nondeterministic and unpredictable strategies to avoid decisions that can be predicted by the intruder. However, existing continuous and adversarial inspection methods are not suitable for dual-task scenarios that simultaneously perform continuous road monitoring and emergency response, but such dual-task requirements are very common in real-world scenarios such as forest fires. Therefore, a dual-task inspection method suitable for continuous road monitoring and emergency response is needed. Summary of the Invention

[0005] This invention provides a dual-task inspection method for continuous road monitoring and emergency response, and applies it to forest fire problems, enabling swarm robots to collaboratively monitor environmental conditions and rapidly handle fires. This dual-task inspection method effectively leverages the practical application potential of swarm robots in responding to changing situations, enabling rapid and effective environmental monitoring and collaborative emergency response.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A dual-task inspection method for continuous road monitoring and handling of emergencies is summarized as follows: Initialize the positions of a swarm of robots; calculate the shortest path between any two nodes, and calculate and store the observable nodes within the monitoring range of each point; update the monitoring status of environmental nodes and the emergencies detected by the swarm of robots at each time step; when a robot reaches an environmental node, it makes a decision on the next movement target, traversing all nodes as candidate nodes; calculate the reward of the candidate nodes and update the shortest path from the robot to them; if the current robot is not in the swarm of robots closest to the emergency, clear its reward so that it will not choose the emergency as its next target; select the target node with the maximum reward for the robot, and the robot moves one step according to the shortest path to approach its target node. Repeat the above decision-making and movement process until the maximum running time is reached.

[0008] A dual-task inspection method for continuous road monitoring and handling of emergencies, characterized by the following specific implementation steps:

[0009] Establish a road network environment G for inspection, set the number of unexpected events H, and initialize algorithm parameters: number of swarm robots R, robot speed v, and robot monitoring capability f. i Robot monitoring range d c Maximum running time T.

[0010] Step 1.1: Randomly initialize the initial positions p of the swarm of robots. i , i∈{1,2,...,R}.

[0011] Step 1.2: Initialize the target points Q = {Q1, Q2, ..., Q...} stored by each robot for other robots. n}

[0012] Step 1.3: Use the Floyd algorithm to calculate the shortest path between any two points, as shown in the following formula:

[0013] D(i,j)=min(D(i,j),D(i,k)+D(j,k))

[0014] Where i and j represent the two endpoints of the shortest path, and k represents the possible intermediate nodes in the path. By iterative calculation, the shortest path between any two points containing multiple path segments can be obtained.

[0015] Step 1.4: Calculate the monitoring range d for each node i. c The set of monitorable nodes within the range O i .

[0016] Step 1.5: Each time step d t Update the sudden events detected by the robot, and update the urgency level of each node i. The formula is as follows:

[0017]

[0018] Where H i S is the number of sudden events occurring at node i. i and These represent whether the non-separable / separable burst events occurring at node i have been handled, respectively, as shown in the following formula:

[0019]

[0020]

[0021] Where [] represents Iverson brackets, the value is 1 when the condition inside is true, and 0 when it is false. r This indicates whether a robot is equipped to handle emergencies. R′ is the set of robots that arrive at node i. For separable emergencies, a robot can simply arrive at the node, handle the emergency, and leave, as in the garbage disposal scenario. When a robot detects discarded garbage, it will fill the garbage bin as much as possible and then leave without waiting for other robots to appear. However, for inseparable emergencies, such as forest fires, multiple robots need to meet to handle the fire source because if the fire is not extinguished in one go, it will regenerate over time.

[0022] Step 1.6: Each time step d t Update the uncertainty u of each node i i (t), the formula is as follows:

[0023]

[0024] in F is the probability that robot r monitors node i. i Let be the probability of all robots monitoring node i, as shown in the following formula:

[0025]

[0026]

[0027] Where d i d is the distance between the robot and node i. c This is the robot's complete monitoring range. L That is the limit of the robot's detection. In d c Nodes within can be completely monitored with probability 1, in d c With d L The nodes within can be partially monitored with a probability of 0 to 1.

[0028] w i t represents the importance of node i. n It is a parameter that normalizes uncertainty into a dimensionless variable. Δt i It is the monitoring time interval of node i, that is, the time interval between the current time t and the last time it was fully monitored t. i The interval is calculated using the following formula:

[0029] Δt i =tt i

[0030] Step 1.7: When the robot reaches a node, it makes the next movement target decision, traverses all nodes as candidate nodes, calculates the dual-task benefit of the candidate nodes, and updates the shortest path from the robot to them. The dual-task benefit formula is as follows:

[0031]

[0032] in It is the dual-task benefit of robot r choosing node i. α represents the uncertainty of observable nodes around node i. α is the dual task priority, ensuring that sudden events are prioritized for robot handling when detected. r The question is whether robot r has the ability to handle unexpected events. p represents the urgency level of node i. c This is the weighting coefficient for path consumption. d min It is the shortest path distance. v is the robot's speed. w i It represents the importance of node i.

[0033] Step 1.8: If the target node does not have any unexpected events and has been selected by other robots, then clear the current robot's earnings. If the target node has unexpected events and the current robot is not among the robots closest to the unexpected events, clear its earnings so that it will not select the unexpected events as its next target.

[0034] Step 1.9: Select the target node where the robot has the greatest benefit, and move one step to approach its target node according to the shortest path.

[0035] The specific process of step 1.7 is as follows:

[0036] Step 1.7.1: Calculate the dual-task priority parameter α to ensure that the robot prioritizes handling unexpected events. The formula is as follows:

[0037] α=max(|O i |)+1

[0038] Among them O i It is the set of monitorable nodes of node i, obtained from step 1.4.

[0039] Step 1.7.2: Update the shortest path from the robot to the candidate target node. When a sudden event occurs at the target node, the destination of the shortest path is the target node. When there is no sudden event, the destination is one of the nodes in the set of nodes that the target node can monitor, because the robot only needs to reach such a node to monitor the target node. The shortest path between any two nodes is the result calculated using the Floyd algorithm in Step 1.3.

[0040] Step 1.7.3: Adaptively adjust the weight coefficient p of path consumption c The formula is as follows:

[0041]

[0042] Where, p m and p n These are the minimum and ordinary weight coefficients, u max This is the maximum allowable uncertainty. When the average uncertainty in the environment is greater than u... max Or when the robot detects a sudden event, p c From p n drop to p m The importance of path loss is reduced.

[0043] Based on steps 1.7.1-1.7.3 above, traverse all candidate nodes and calculate the dual-task benefit of robot r choosing node i.

[0044] The specific process of step 1.8 is as follows:

[0045] Step 1.8.1: If the target node does not have any sudden events and is selected by other robots, then the current robot's reward for selecting the node is 0.

[0046] Step 1.8.2: Obtain the target node Q of other robots, calculate the number of times k that the current robot's target viewpoint is the same as the node in Q, and add the record in Q to the waiting list, which stores the robot index with the same target and its reward.

[0047] Step 1.8.3: If k = H i The currently selected node already has enough robots assigned to handle it. Therefore, the current robot needs to compete with the robot in Q that has the lowest payoff. If the competition is successful, the robot originally in Q is removed. If the competition is unsuccessful, the current robot's payoff for choosing this node is 0. After all robots have made their decisions, the group of robots closest to the event will autonomously choose the target node to handle the event.

[0048] Based on steps 1.8.1-1.8.3 above, if the current robot is not among the group of robots closest to the emergency, it will not choose the emergency as its next target.

[0049] The dual-task inspection method proposed in this invention enables a group of robots to efficiently and collaboratively monitor and quickly handle emergencies such as fires. Attached Figure Description

[0050] Figure 1 This is a flowchart of the dual-task inspection method of the present invention.

[0051] Figure 2 The flowchart for step 1.7 of the dual-task inspection method of the present invention is shown.

[0052] Figure 3 The flowchart for step 1.8 of the dual-task inspection method of the present invention is shown.

[0053] Figure 4 This is a schematic diagram of a dual-task forest fire scenario in an embodiment of the present invention.

[0054] Figure 5 These are three different road maps according to embodiments of the present invention.

[0055] Figure 6 This is a schematic diagram of the motion trajectory of the swarm robot according to an embodiment of the present invention.

[0056] Figure 7 This is a convergence curve of risk level in a dual-task scenario of forest fire handling according to an embodiment of the present invention. Detailed Implementation

[0057] Application of the described dual-task inspection method in forest fire problems:

[0058] In forest fire management issues (such as...) Figure 4 As shown, this dual-task inspection method is applied to the monitoring and handling of fire sources. Assuming each fire source is an inseparable, sudden event, robots are deployed within a road network and move along the roads. When a robot fails to detect a fire source, it continuously monitors every node in the road network to ensure the overall environmental uncertainty remains low, thus avoiding overlooking potential fire sources. During continuous monitoring, once any robot detects a fire source, it shares its location within the group. The robot swarm then collaboratively allocates the nearest group of robots to handle the fire source first. Since a fire source may require multiple robots to extinguish, a robot waiting for others to arrive will wait for its turn. Once the fire source is extinguished, the robot swarm continues to monitor the environment to detect other potential fire sources. This process is entirely autonomous and requires no manual task phase division.

[0059] The implementation scheme adopted in this embodiment of the invention is as follows:

[0060] The process flow of a dual-task inspection method for continuous road monitoring and handling of emergencies is as follows: Figure 1 As shown.

[0061] initialization:

[0062] Establish a road network environment G for inspection, set the number of unexpected events H, and initialize algorithm parameters: number of swarm robots R, robot speed v, and robot monitoring capability f. i Robot monitoring range d c Maximum running time T.

[0063] Preparation steps:

[0064] Step 1.1: Randomly initialize the initial positions p of the swarm of robots. i , i∈{1,2,...,R}.

[0065] Step 1.2: Initialize the target points Q = {Q1, Q2, ..., Q...} stored by each robot for other robots. n}

[0066] Step 1.3: Use the Floyd algorithm to calculate the shortest path between any two points, as shown in the following formula:

[0067] D(i,j)=min(D(i,j),D(i,k)+D(j,k))

[0068] Where i and j represent the two endpoints of the shortest path, and k represents the possible intermediate nodes in the path. By iterative calculation, the shortest path between any two points containing multiple path segments can be obtained.

[0069] Step 1.4: Calculate the monitoring range d for each node i. c The set of monitorable nodes within the range O i .

[0070] At each time step, update the environmental state and determine whether the robot has reached the road node:

[0071] Step 1.5: Each time step d t Update the fire sources detected by the robot, and update the urgency level of each node i. The formula is as follows:

[0072]

[0073] Where H i S is the number of robots required to extinguish the fire at node i. i The formula indicating whether the fire source at node i has been dealt with is as follows:

[0074]

[0075] Where [] represents Iverson brackets, the value is 1 when the condition inside is true, and 0 when it is false. r This indicates whether a robot has a fire extinguisher to deal with a fire. R′ is the set of robots that reach node i.

[0076] Step 1.6: Each time step d t Update the uncertainty u of each node i i (t), the formula is as follows:

[0077]

[0078] in F is the probability that robot r monitors node i. i Let be the probability of all robots monitoring node i, as shown in the following formula:

[0079]

[0080]

[0081] Where d i d is the distance between the robot and node i. c This is the robot's complete monitoring range. L That is the limit of the robot's detection. In d c Nodes within can be completely monitored with probability 1, in d c With d L The nodes within can be partially monitored with a probability of 0 to 1.

[0082] w i t represents the importance of node i. n It is a parameter that normalizes uncertainty into a dimensionless variable. Δt i It is the monitoring time interval of node i, that is, the time interval between the current time t and the last time it was fully monitored t. i The interval is calculated using the following formula:

[0083] Δt i =tt i

[0084] Iteration and inspection decisions:

[0085] Step 1.7: When the robot reaches a node, it makes the next movement target decision, traverses all nodes as candidate nodes, calculates the dual-task benefit of the candidate nodes, and updates the shortest path from the robot to them. The dual-task benefit formula is as follows:

[0086]

[0087] in It is the dual-task benefit of robot r choosing node i. α represents the uncertainty of observable nodes around node i. α is the dual task priority, ensuring that the fire source is prioritized for robot processing when detected. r The question is whether robot r is capable of handling fire sources. p represents the urgency level of node i. c This is the weighting coefficient for path consumption. d min It is the shortest path distance. v is the robot's speed. w i It represents the importance of node i.

[0088] Step 1.8: If the target node does not have a fire source and has been selected by another robot, then clear the current robot's earnings. If the target node has a fire source and the current robot is not in the group of robots closest to the fire source, clear its earnings so that it will not choose the fire source as its next target.

[0089] Step 1.9: Select the target node where the robot has the greatest benefit, and move one step to approach its target node according to the shortest path.

[0090] like Figure 2 As shown, the specific process of step 1.7 is as follows:

[0091] Step 1.7.1: Calculate the dual-task priority parameter α to ensure the robot prioritizes handling the fire source. The formula is as follows:

[0092] α=max(|O i |)+1

[0093] Among them O i It is the set of monitorable nodes of node i, obtained from step 1.4.

[0094] Step 1.7.2: Update the shortest path from the robot to the candidate target node. When a sudden event occurs at the target node, the destination of the shortest path is the target node. When there is no source of fire, the destination is one of the nodes in the set of nodes that the target node can monitor, because the robot only needs to reach such a node to monitor the target node. The shortest path between any two nodes is the result calculated using the Floyd algorithm in Step 1.3.

[0095] Step 1.7.3: Adaptively adjust the weight coefficient p of path consumption c The formula is as follows:

[0096]

[0097] Where, pm and p n These are the minimum and ordinary weight coefficients, u max This is the maximum allowable uncertainty. When the average uncertainty in the environment is greater than u... max Or when the robot detects a fire source, p c From p n drop to p m The importance of path loss is reduced.

[0098] Based on steps 1.7.1-1.7.3 above, traverse all candidate nodes and calculate the dual-task benefit of robot r choosing node i.

[0099] like Figure 3 As shown, the specific process of step 1.8 is as follows:

[0100] Step 1.8.1: If the target node does not have a fire source and is selected by other robots, the current robot's reward for selecting the node is 0.

[0101] Step 1.8.2: Obtain the target node Q of other robots, calculate the number of times k that the current robot's target viewpoint is the same as the node in Q, and add the record in Q to the waiting list, which stores the robot index with the same target and its reward.

[0102] Step 1.8.3: If k = H i The currently selected node already has enough robots assigned to handle it. Therefore, the current robot needs to compete with the robot in Q that has the lowest reward. If the competition is successful, the robot in Q is removed. If the competition is unsuccessful, the current robot's reward for choosing this node is 0. After all robots have made their decisions, the group of robots closest to the fire source will autonomously choose the target node to handle the fire.

[0103] Based on steps 1.8.1-1.8.3 above, if the current robot is not among the group of robots closest to the fire source, it will not choose the fire source as its next target.

[0104] To more intuitively verify the performance of this invention in forest fire management, the embodiments used three types of maps—small-scale, irregular, and large-scale—to simulate complex road network environments, as shown below. Figure 5 (a) Figure 5 (b) Figure 5 As shown in (c), fire sources appear randomly at road network nodes in these maps at a certain frequency. The algorithm's performance is evaluated by the average risk level of all nodes (C1) and the percentage of successfully handled fire sources (C2), as shown in the following formula:

[0105]

[0106] C2=H s / H t

[0107] Among them, u i (t) is the uncertainty of node i at time t, which can be obtained from step 1.6. H represents the urgency of node i at time t, which can be obtained from step 1.5. s H represents the number of fire sources successfully dealt with. t It represents the total number of fire sources.

[0108] The operating effect of this invention on irregular maps is as follows: Figure 6 and Figure 7 As shown, initially, the robots continuously monitor nodes in the environment, causing the risk level to decrease. When a fire source appears, the environmental risk level suddenly increases. At this point, the swarm of robots detects the fire source first and then deals with it with higher priority. Once the fire source is dealt with, the risk level drops to a lower level, and the swarm of robots continues to monitor environmental nodes until a new fire source appears. As the monitoring range increases, the robots can detect fire sources more quickly, and the risk level decreases in a shorter time. The above results and analysis demonstrate the feasibility of this invention in handling the dual tasks of forest fires.

[0109] The dual-task patrol method (CDTPP) proposed in this invention for continuous road monitoring and handling of emergencies is compared with the most outstanding heuristic cooperative patrol algorithm (HCPS) (T. Wang, P. Huang, and G. Dong, “Cooperative persistent surveillance on a road network by multi-ug vs with detection ability,” IEEE Transactions on Industrial Electronics, vol. 69, no. 11, pp. 11468-11478, Nov. 2022.) and its improved version (HCPS-m). The results are shown in Table I. CDTPP has a lower risk level and a higher fire source handling rate compared with HCPS and HCPS-m. This shows that using the CDTPP method can significantly improve the ability of swarm robots to monitor and collaboratively handle emergencies such as fire sources, thereby quickly and effectively extinguishing fires.

[0110] Table I

[0111] The proposed dual-task inspection method (CDTPP) is compared with HCPS and HCPS-m under different numbers of robots (C1 / C2).

[0112]

[0113] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A dual-task patrol method for road continuous monitoring and handling of incidents, characterized in that, The implementation steps are: Initialization: Establish a patrol road network environment G, set the number of emerging incidents H; and then Initialize algorithm parameters: number of robots in the group R, robot speed v, robot monitoring ability f i , robot monitoring range d c , maximum running time T; Preparation steps: Step 1.1: Random initial population of robots in the road network environment G at positions p i i e {1,2,...,R}; Step 1.2: Initialize each robot stored target points of other robots Q = {Q1, Q2,..., QN} for i = 1 to N do n} Step 1.3: Use the Floyd algorithm to calculate the shortest path between any two points in the road network environment G, the formula is as follows: D(i,j) = min(D(i,j), D(i,k) + D(j,k)) Where i, j represent the two end nodes of the shortest path, and k represents the possible intermediate node in the path. Through iterative calculation, the shortest path containing multiple segments between any two points is obtained; Step 1.4: Calculate the set of monitorable nodes O within the monitoring range d for each node i c i ;​ Update the environment state at each time step and determine whether the robot has reached the road node: Step 1.5: At each time step d t Update the emergency level of each node i The formula is as follows: where H i is the number of incidents that occurred at node i; S i and respectively indicate whether the non-separable / separable incidents that occurred at node i were handled, as follows: where [] is the Iverson bracket, which takes the value 1 when the condition inside is true and 0 when it is false; I r represents whether a robot is equipped to handle an emergency; R' is the set of robots that arrive at node i; for separable emergencies, a robot can arrive at a node, handle an emergency, and leave; while inseparable emergencies, for the forest fire scenario, require multiple robots to meet to handle the fire source; Step 1.6: Each time step d t Update the uncertainty u of each node i i (t), as follows: where f i r is the probability of robot r monitoring node i, F i is the probability of all robots monitoring node i, given by the formula: where d i is the distance of the robot from node i; d c is the fully monitorable range of the robot; d L is the detection limit of the robot; nodes within d c are fully monitored with probability 1, nodes within d c are partially monitored with probability 0-1; and d L is the partially monitorable range of the robot. w i is the importance degree of node i; t n is a parameter that normalizes the uncertainty to a dimensionless variable; Δt i is the monitoring time interval of node i, i.e. the interval between the current time t and the last time t i that node i was fully monitored, and is given by the following formula: Δt i = t - t i Iteration and patrol decision: Step 1.7: When the robot reaches a node, make the next move target decision, traverse all nodes as candidate nodes, calculate the double-task benefit of the candidate nodes and update the shortest path of the robot to them, the double-task benefit formula is as follows: wherein is the dual-task reward of robot r selecting node i; is the uncertainty of observable nodes around node i; a is the dual-task priority, ensuring that the emergency can be handled by the robot when it is monitored; I r is whether the robot r has the ability to handle the emergency; is the urgency of node i; p c is the weight coefficient of path consumption; d min is the shortest path distance; v is the robot speed; w i is the importance of node i; Step 1.8: If the target node does not have an incident and is selected by other robots, clear the current robot's benefit; if the target node has an incident and the current robot is not in the group of robots closest to the incident, clear its benefit so that it will not choose the incident as the next target; Step 1.9: Select the target node with the largest benefit for the robot, and the robot moves one step according to the shortest path to approach its target node.

2. The dual task inspection method for road continuous monitoring and handling of emergent events as claimed in claim 1 wherein, In step 1.7, the double-task benefit of the candidate nodes is calculated and the shortest path of the robot to them is updated, the specific process of step 1.7 is as follows: Step 1.7.1: Calculate the double-task priority parameter a to ensure that the robot prioritizes handling incidents, the formula is as follows: a = max (|O i |) + 1 where O i is the set of monitorable nodes of node i, obtained by step 1.4; Step 1.7.2: Update the shortest path of the robot to the candidate target nodes; when the target node has an incident, the destination of the shortest path is the target node; when there is no incident, the destination is one of the nodes in the target node's monitorable node set, because the robot only needs to reach such a node to monitor the target node; the shortest path between any two nodes is the result calculated using the Floyd algorithm in step 1.3; Step 1.7.3: Adaptive adjustment of the weight coefficient p of the path consumption c The formula is as follows: where p m and p n are minimum and ordinary weight coefficients, respectively, u max is the maximum value of uncertainty allowed; when the average uncertainty in the environment is greater than u max or the robot detects an emergency, p c decreases from p n to p m , and the importance of path loss decreases; Based on the above steps 1.7.1-1.7.3, all candidate nodes are traversed and the double-task revenue of the robot r selecting the node i is calculated 3. The dual task inspection method for road continuous monitoring and handling of emergent events as claimed in claim 1 wherein, In step 1.8, the robot's repeated benefits are cleared, the specific process of step 1.8 is as follows: Step 1.8.1: If the target node does not have an incident and is selected by other robots, the current robot's benefit for selecting this node is 0; Step 1.8.2: Obtain the target nodes Q of other robots, calculate the number of times k that the current robot's target view is the same as the nodes in Q, and add the record in Q to the waiting list, which stores the indices of robots with the same target and their benefits; Step 1.8.3: If k = H i , the currently selected node has allocated enough robots to handle it; therefore the current robot needs to compete with the robot in Q with the lowest payoff; if the competition is successful, the robot in Q is removed; If the competition fails, the current robot's benefit for selecting this node is 0; after all robots have made a decision, the group of robots closest to the incident will autonomously select the target node to handle the incident; Based on the above steps 1.8.1-1.8.3, if the current robot is not in the group of robots closest to the incident, it will not choose the incident as the next target.

Images