Multi-robot control system

Abstract

The invention discloses a multi-robot control system, and relates to the field of robot path planning. According to the system, in a robot path driving task, candidate path nodes are selected by randomly switching target guidance and a global exploration mode, feasible nodes and path segments are determined by combining historical path nodes, and the path nodes are expanded after collision detection of a static obstacle and a dynamic robot. The system also dynamically adjusts a driving step length based on a load rate, sets a safety margin including load compensation, and realizes multi-machine cooperation through local path topological optimization and distributed decision. According to the method, the multi-robot path conflict and collision risk is effectively reduced, the heavy-load working condition motion stability and the system cooperation efficiency are improved, and the method is suitable for a multi-robot cooperation operation scene.

Description

Technical Field

[0001] The present invention relates to the field of robot path planning, in particular to a multi-robot control system. Background Art

[0002] In path planning scenarios for heavy-loaded machinery in industrial plants, high path conflicts and collision risks are primary challenges. Traditional RRT algorithms employ a global uniform random sampling strategy, making it impossible to effectively avoid path intersections when multiple heavy-loaded vehicles are operating simultaneously. This is particularly true in obstacle-heavy plant environments, where accidents such as equipment collisions and cargo drops can easily occur, severely threatening operational safety and leading to production line disruptions. Furthermore, existing collision detection methods ignore the robot's size and load characteristics, relying solely on basic bounding box detection. This results in significant blind spots, further exacerbating the risk of static obstacles colliding with the dynamic robot.

[0003] The core flaw of traditional technologies is the lack of stability under heavy loads. The fixed extended step length mechanism cannot adapt to the varying loads of heavily loaded carts. When the cart is carrying heavy objects, the increased inertia causes vibration amplitudes far exceeding safety standards during startup, acceleration, and cornering. This not only affects the lifespan of the machine and the safety of the cargo (such as damage to precision components due to vibration), but can also cause loss of control due to the fixed step length on complex terrain (such as uneven ground). For example, a fixed step length under heavy loads can exacerbate the center of gravity shift of the cart, significantly increasing the risk of rollover. However, when the step length is lightly loaded, failure to increase in time reduces path planning efficiency.

[0004] The low efficiency of multi-machine collaboration and limited system scalability have hampered factory operations. Traditional centralized scheduling relies on a central controller to process massive amounts of real-time data, which is prone to information lags and decision-making delays, and cannot meet the needs of real-time obstacle avoidance and task collaboration for multiple heavy-loaded vehicles. When new equipment is added to the factory or the operation process is adjusted, the centralized system needs to modify parameters and programs on a large scale, and its scalability is extremely poor. In addition, the lack of a distributed collaboration mechanism prevents robots from sharing environmental information in real time, and the dynamic optimization capability of paths is insufficient. Local congestion often leads to a decrease in global efficiency, making it difficult to adapt to the complex scenarios in the factory where multiple tasks are carried out in parallel and dynamic obstacles frequently change. Summary of the Invention

[0005] The present invention aims to solve the above-mentioned problems.

[0006] To this end, the technical solution adopted in the present invention is as follows:

[0007] A multi-robot control system that implements the following strategies when the robots perform path driving tasks:

[0008] By randomly switching between the goal-oriented mode and the global exploration mode, candidate path nodes are selected;

[0009] Selecting the robot's historical path node closest to the candidate path node as a feasible node;

[0010] Return the optimal parent node among the feasible nodes and the path segments between the candidate path nodes and the optimal parent node;

[0011] Detecting whether the robot collides with static obstacles or other robots in the path segment;

[0012] If the test passes, the optimal parent node is used as the new expansion node.

[0013] Furthermore, the robot's driving step length s is adjusted by the basic step length and the scaling factor; the formula is as follows:

[0014]

[0015] in, is the basic step length, K is the minimum step length;

[0016] The scaling factor k is determined based on the nonlinear attenuation formula and the robot load rate, as follows:

[0017]

[0018] in, is the weight coefficient, The load rate is the ratio of the actual load to the maximum load.

[0019] Furthermore, the safety margin of the robot's travel is set, which is composed of the robot's physical width, the load dynamic compensation term, and the fixed redundancy value. The formula is as follows:

[0020]

[0021] in, is the basic safety margin, , width is the physical width of the robot, is the load dynamic compensation term, , is a fixed redundancy value.

[0022] Furthermore, the generation process of the candidate path nodes is as follows:

[0023] Set the target sampling probability threshold and generate a random number in the range [0, 100];

[0024] When the random number is greater than the target sampling probability threshold, the target bias mode is executed, and the steps are as follows:

[0025] Extract the end coordinates of the robot's historical path node sequence to obtain the robot's current node position coordinates; calculate the Euclidean vector of the target node position and the current node position to obtain the target direction vector; use the four-quadrant inverse tangent function to calculate the target direction angle θ; generate a uniformly distributed random sampling distance , the formula is as follows:

[0026]

[0027] Among them, ||V|| is the straight-line distance from the current node position to the target node position, and 1.2 is the expansion coefficient;

[0028] A random deviation angle is superimposed on the target direction angle to generate a random sampling angle. , the formula is as follows:

[0029]

[0030] in, , is the uniform distribution function;

[0031] By the random sampling distance and the random sampling angle , obtain new random sampling points as the candidate path nodes;

[0032] When the random number is not greater than the target sampling probability threshold, a global exploration mode is executed to uniformly randomly sample the entire drivable space to obtain a new random sampling point as the candidate path node.

[0033] Furthermore, the collision detection process is as follows:

[0034] Receive the endpoint coordinates (x1, y1) to (x2, y2) of the path segment to be detected, and the current robot index identifier;

[0035] Enter the static detection phase, traverse the obstacle list, and perform the following sub-process for each obstacle:

[0036] An axis-aligned bounding box fast detection algorithm is used to calculate the minimum Euclidean distance between the path segment and the obstacle as the shortest path distance;

[0037] Perform static collision judgment. The static collision condition is whether the square of the shortest path distance is greater than the square of the sum of the obstacle radius and the robot body radius. If so, return False; otherwise, return True.

[0038] Enter the dynamic detection phase, traverse the robot list, and perform the following sub-process for each robot:

[0039] When the traversed robot list index is equal to the current robot index identifier, the dynamic detection phase is skipped;

[0040] Otherwise, determine the following dynamic collision conditions:

[0041]

[0042] Where path(t) is the linear interpolation position of the path segment, robot(t) is the predicted trajectory position of the robot; robot is the radius of the robot body;

[0043] If the dynamic collision condition is not met, it returns False, otherwise it returns True;

[0044] If True is returned, it means a collision is detected;

[0045] If False is returned, it indicates that the path segment is safe, and the corresponding candidate path node passes the test and serves as the new extended node.

[0046] Furthermore, after obtaining the new expansion node, an optimization radius is defined with the new expansion node as the center, all historical path nodes of the robot within the optimization radius are traversed, and an optimal path from the current position node to the new expansion node is generated as an alternative path;

[0047] The robot's historical path node sequence and the parent relationship of all nodes in the robot's historical path node sequence are updated.

[0048] Furthermore, after completing a single path segment, the robot performs a status check, including but not limited to: calculating the Euclidean distance between the current position node and the target position node,

[0049] If the Euclidean distance is less than the specified distance and the velocity vector converges, the driving task is marked as completed, the robot resources are released and the system status list is updated; otherwise, the driving task is marked as incomplete and the robot continues to enter the next driving task strategy execution cycle.

[0050] Compared with the prior art, the advantages of the present invention are:

[0051] (1) The present invention generates candidate paths by dynamically switching between the goal-oriented mode and the global exploration mode. Combined with the dual-modal collision detection mechanism, it expands the exploration range while ensuring the efficiency of path search, effectively reduces the probability of multi-robot path intersection conflicts, and improves the safety of collaborative operations.

[0052] (2) The present invention dynamically adjusts the driving step length based on the nonlinear attenuation formula of the load rate, and adopts a multi-level safety margin protection mechanism to ensure that the system significantly improves the motion stability under heavy load conditions, avoids the vibration problem caused by fixed step length, and ensures the safety of cargo transportation.

[0053] (3) The present invention adopts local path topology optimization and distributed decision-making mechanism. The system realizes real-time environmental information sharing and dynamic path optimization among multiple robots without relying on centralized scheduling, thereby improving resource utilization and task execution efficiency, and enhancing system scalability and flexibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0055] Figure 1 It is the overall flow chart of the present invention;

[0056] Figure 2 This is a random sampling flow chart of the present invention;

[0057] Figure 3 This is a collision detection flow chart of the present invention;

[0058] Figure 4 This is a schematic diagram of load test parameters of the present invention;

[0059] Figure 5 This is a diagram of the visualization process when the robot load rate is 0.2;

[0060] Figure 6 This is a diagram of the visualization process when the robot load rate is 0.4;

[0061] Figure 7 This is a diagram of the visualization process when the robot load rate is 0.6;

[0062] Figure 8 This is a diagram of the visualization process when the robot load rate is 0.8;

[0063] Figure 9 Schematic diagram of the visualization process when the robot load rate is 1. DETAILED DESCRIPTION

[0064] To achieve the above objectives, the present invention is implemented through the following technical solutions. The present invention provides a multi-robot control system, such as Figure 1 As shown;

[0065] The system continuously monitors the activity of all robots. If a robot fails to complete its path-following task, a multi-robot collaborative planning loop is initiated. Each robot independently executes the following six-stage drive plan, and the central controller achieves collaborative decision-making by sharing environmental information.

[0066] Phase 1: Dynamic Step Size Adjustment

[0067] Read the current load rate of the robot, which is equal to the ratio of the actual load to the maximum load;

[0068] The robot's step length scaling factor is calculated by the load rate and the nonlinear attenuation formula. The nonlinear attenuation formula in this embodiment is a negative exponential function: , where k is the scaling factor, is the weight coefficient, is the load rate.

[0069] The dynamic step length formula of the robot is: ,in is the basic step length, K is the minimum step length;

[0070] In this embodiment, , , .

[0071] This method limits the robot's travel step length to a preset safety range, automatically reducing the travel step length under heavy load conditions to 15% to 20% of that under light load conditions, ensuring smooth movement.

[0072] At the same time, set the safe distance of the robot during driving, that is, the safety margin;

[0073] The safety margin is composed of the robot's physical width, load dynamic compensation terms, and fixed redundancy values, forming a multi-level protection mechanism. The formula is as follows:

[0074]

[0075] in, is the basic safety margin, , width is the physical width of the robot, is the load dynamic compensation term, , is a fixed redundancy value.

[0076] In this embodiment, , , the output safety margin is between [0.1,0.3].

[0077] In this embodiment, Phase 1 achieves dynamic adjustment of motion parameters through real-time load sensing, including two core functions: step-size adaptive control and safety margin generation. Based on the normalized load rate, a negative exponential function with a damping coefficient in the denominator is used to calculate the step-size scaling factor, compressing the step size to a preset safety range to ensure motion smoothness under heavy load conditions. At the same time, a safety margin model consisting of the robot's physical width, load dynamic compensation terms, and fixed redundancy values ​​is constructed to form a multi-level protection mechanism. The technical solution achieves dynamic matching of load parameters and motion control through decoupling design, automatically balancing path planning efficiency and motion safety within a given load range.

[0078] Phase 2: Goal-directed sampling

[0079] A dual-mode random sampling strategy is used to generate candidate path nodes. This strategy takes into account both path search efficiency and global optimality, and reduces the number of invalid sampling. They are as follows: Figure 2 As shown:

[0080] First, target bias mode (trigger probability 15%)

[0081] Second, the global exploration mode (trigger probability 85%) performs uniform random sampling across the entire drivable space.

[0082] The execution process of the target bias mode is as follows:

[0083] The vector V pointing to the target position is obtained by determining the current position C and the target position G. Rate represents the probability threshold (percentage) of the target being selected. When the generated random number exceeds the probability threshold, the target bias mode is executed, otherwise the global exploration mode is executed. The target direction angle is obtained by the inverse trigonometric function. , get the random sampling distance r, and finally according to r and target direction angle Sampling. The specific steps are as follows:

[0084] Step 1: Set the target sampling probability threshold Rate and generate a random number in the range of 0-100. When the random number is greater than the target sampling probability threshold, execute the global exploration mode.

[0085] Step 2: Extract the end coordinates of the robot's historical path node sequence to obtain the robot's current node position coordinates;

[0086] Step 3: Calculate the Euclidean vector of the target node position and the current node position to obtain the target direction vector;

[0087] Step 4: Use the four-quadrant inverse tangent function to calculate the target direction angle θ, ensuring that the angle value is in the correct quadrant [-π, π];

[0088] Step 5: Generate a uniformly distributed random sampling distance: r ~ U(0, 1.2×||V||), where ||V|| is the straight-line distance from the current node to the target node, and 1.2 is the expansion factor used to enhance the coverage of the sampling area.

[0089] Step 6: Superimpose the random deviation angle on the target direction angle to generate a random sampling angle: φ = θ + δ, where δ ~ U(-π / 4, π / 4);

[0090] Obtain new random sampling points through random sampling distance and random sampling angle;

[0091] Step 7: Convert the new random sampling points into candidate path node coordinates through polar coordinate transformation.

[0092] Phase 3: Neighbor Node Retrieval

[0093] A spatial hierarchical index of the robot's historical path nodes is established, and the branch-and-bound method is used to quickly locate the N robot historical path nodes closest to the candidate path node as feasible nodes. The index of the optimal feasible node among the feasible nodes and the path segment between the random sampling point and the optimal parent node are returned.

[0094] The above process also uses KD trees to accelerate the search speed. Phase 4: Bimodal Collision Detection Bimodal collision detection is performed on the path segments, including a static detection phase and a dynamic detection phase, respectively, to handle the real-time collision risk of static obstacles and dynamic robots. Through fast geometric calculations based on Euclidean distance, step-by-step detection is performed in the obstacle list and robot list, where:

[0095] The static detection stage uses the square calculation of the point-line segment distance to avoid sqrt operations and improve efficiency;

[0096] In the dynamic detection phase, self-collision elimination is achieved through index filtering;

[0097] The following is a detailed introduction to Phase 4. Figure 3 As shown:

[0098] Step 1: Receive the coordinates of the endpoints of the path segment to be detected (x1, y1) to (x2, y2), as well as the current robot index identifier;

[0099] Step 2: Enter the static detection phase, traverse the obstacle list and perform the following sub-steps:

[0100] Step 2.1: Use the axis-aligned bounding box fast detection algorithm to calculate the minimum Euclidean distance between the path segment and the obstacle as the shortest path distance;

[0101] Step 2.2: Perform static collision judgment. The static collision condition is whether the square of the shortest path is greater than the square of the sum of the obstacle radius and the robot body radius. If so, return False, otherwise return True;

[0102] Step 3: Enter the dynamic detection phase, traverse the robot list and perform the following sub-steps:

[0103] Step 3.1 Self-detection mechanism: When the traversed robot list index is equal to the current robot index identifier, the dynamic detection phase is skipped;

[0104] Step 3.2: Perform dynamic collision judgment. The dynamic collision conditions are:

[0105] robot(t)−path(t)≤robot,∃t∈[0,1]

[0106] Where path(t) is the linear interpolation position of the path segment, robot(t) is the predicted trajectory position of the robot, and robot is the radius of the robot body. If the dynamic collision condition is not met, it returns False, otherwise it returns True.

[0107] Step 4: If False is returned, it means a collision is detected, and the collision type is static or dynamic; if True is returned, it means the path segment is safe, and the corresponding candidate path node passes the detection and serves as a new extension node.

[0108] Phase 5: Path Topology Optimization

[0109] An optimization radius is defined with the expansion node as the center, all historical path nodes within the radius are traversed, and the optimal path from the current position node to the expansion node is generated as an alternative path and the parent relationship of all nodes and the robot's historical path sequence are updated.

[0110] Phase 6: Task Status Update After completing a single path segment, a status check is performed, including:

[0111] Calculate the Euclidean distance between the current location node and the target location node,

[0112] If the distance is less than the specified distance and the velocity vector converges, the task is marked as completed, the robot resources are released, and the system status list is updated; otherwise, the task is marked as incomplete and the robot continues to enter the next driving planning cycle.

[0113] The performance test experiments of the present invention are as follows:

[0114] Experiment 1: Dynamic load adaptation test: Evaluate the total path length and average planning time of a single robot after dynamic load adaptive adjustment of the step size under different load conditions.

[0115] Setting the robot load factor in simulation From no load to full load, the parameters are divided into five groups with a decrease interval of 0.2. Figure 4 shown.

[0116] Figure 4 The first row of parameters in the first group is the maximum load of 50, the second row is the number of robots, 1, and the third row is the horizontal coordinate of 0, the vertical coordinate of 0, and the load of 10. The load factor of the first group is 10 / 50 = 0.2. The parameters of the second to fifth groups have the same meaning as the first group. The calculated load factors are 0.4, 0.6, 0.8, and 1.0, respectively.

[0117] The simulation visualization process of the five groups is shown in Figure 5-Figure 9 .

[0118] After the experiment is over, the total length of the robot's planned path is calculated. The length of the final path obtained by performing path planning on the above five sets of data will be recorded.

[0119] At the same time, the average time taken by the robots to plan their paths is calculated. Each robot group records the time it took to plan the path each time it finds its destination. The total time required for the robot's path planning and the number of successful attempts in the current planning process are then calculated, ultimately yielding the total time it took to successfully plan a path.

[0120] The statistical results are shown in Table 1:

[0121] Table 1 The impact of different loads on algorithm path planning

[0122] #timg# Average planning time (s) Path length (m) 0.2 6.64 19.34 0.4 9.23 19.36 0.6 9.83 19.35 0.8 10.07 19.31 1 10.36 19.25

[0123] From the test results, it can be seen that as the load of the heavy-duty robot continues to increase, although the average path-finding time increases to a certain extent, the total planned path length remains basically unchanged, which reflects the stability of the algorithm proposed in this invention.

[0124] In addition, when the load ratio is 0.4, the present invention is compared with the RRT, RRT* and Informed-RRT* algorithms.

[0125] The RRT algorithm is a sampling-based motion planning algorithm that randomly samples points in a state space (such as a robot's workspace) and gradually constructs a tree rooted at the starting point. The tree explores the space to find a path from the starting point to the target point. It efficiently searches for paths in high-dimensional spaces or complex environments (with obstacles) without relying on a precise analytical model of the environment.

[0126] RRT* is an improved version of RRT. Its core is to optimize the tree through rewiring operations during the construction of random trees, so that the final path is closer to the optimal one. It is often used in scenarios such as robot path planning and autonomous driving path search.

[0127] Informed-RRT* is a further improvement on RRT*, introducing heuristic information to guide sampling points to be more concentrated in areas where more optimal paths may exist. This speeds up the algorithm's convergence to the optimal path, reduces unnecessary sampling exploration, improves planning efficiency and path quality, and is widely used in scenarios such as high-precision robot path planning.

[0128] As can be seen from Table 2, the path planning efficiency of the Informed-RRT* and RRT* algorithms is far lower than that of the improved algorithm proposed in this paper. Although the RRT algorithm has a shorter average time, according to the characteristics of the RRT algorithm, as the map complexity continues to increase, the time advantage of RRT will be diluted by the randomly extended path.

[0129] Table 2 Comparison of efficiency of different algorithms under load rate 0.4

[0130] algorithm Average planning time (s) Path length (m) Normalized score RRT 5.63 22.0 0.400 RRT* 10.51 19.41 0.822 Informed-RRT* 18.56 19.50 0.568 The present invention 9.23 19.36 0.888

[0131] The weight used for the normalized scoring is: average planning time (s)*0.6+path length (m)*0.4.

[0132] Experiment 2: Multi-machine density stress test: Increase the number of collaborative robots to verify whether the algorithm can operate normally in a high-density environment.

[0133] Table 3 Comparison of efficiency of different algorithms under multi-robot collaboration

[0134] algorithm Average planning time (s) Path length (m) RRT 17.81 64.63 RRT* 46.81 46.76 Informed-RRT* 35.02 34.04 The present invention 34.69 32.44

[0135] This experiment evaluated the total path length and average planning time of three robots using different algorithms under collaborative operation. The load factors of the three robots were 0.2, 0.3, and 0.4, respectively. The results are shown in Table 3. Although the average planning time was still longer than that of the RRT algorithm, the proposed method surpassed the other algorithms in terms of path length. In particular, the path length of the RRT algorithm, when working with multiple robots, increased almost exponentially with the number of robots.

[0136] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A multi-robot control system, characterized in that: When the robot performs a path driving task, it executes the following strategies: By randomly switching between the goal-oriented mode and the global exploration mode, candidate path nodes are selected; Selecting the robot's historical path node closest to the candidate path node as a feasible node; Return the optimal parent node among the feasible nodes and the path segments between the candidate path nodes and the optimal parent node; Detecting whether the robot collides with static obstacles or other robots in the path segment; If the test is passed, the optimal parent node is used as the new expansion node.

2. The system according to claim 1, wherein: The robot's driving step length s is adjusted by the basic step length and the scaling factor; the formula is as follows: in, is the basic step length, K is the minimum step length; The scaling factor k is determined based on the nonlinear attenuation formula and the robot load rate, as follows: in, is the weight coefficient, The load rate is the ratio of the actual load to the maximum load.

3. The system according to claim 1, wherein: Set the robot's travel safety margin, which is composed of the robot's physical width, load dynamic compensation term, and fixed redundancy value. The formula is as follows: in, is the basic safety margin, , width is the physical width of the robot, is the load dynamic compensation term, , is a fixed redundancy value.

4. The system according to claim 1, wherein: The generation process of the candidate path nodes is as follows: Set the target sampling probability threshold and generate a random number in the range [0, 100]; When the random number is greater than the target sampling probability threshold, the target bias mode is executed, and the steps are as follows: Extract the end coordinates of the robot's historical path node sequence to obtain the robot's current node position coordinates; calculate the Euclidean vector of the target node position and the current node position to obtain the target direction vector; use the four-quadrant inverse tangent function to calculate the target direction angle θ; generate a uniformly distributed random sampling distance , the formula is as follows: Among them, ||V|| is the straight-line distance from the current node position to the target node position, and 1.2 is the expansion coefficient; A random deviation angle is superimposed on the target direction angle to generate a random sampling angle. , the formula is as follows: in, , is the uniform distribution function; By the random sampling distance and the random sampling angle , obtain new random sampling points as the candidate path nodes; When the random number is not greater than the target sampling probability threshold, a global exploration mode is executed to uniformly randomly sample the entire drivable space to obtain a new random sampling point as the candidate path node.

5. The system according to claim 1, wherein: The collision detection process is as follows: Receive the endpoint coordinates (x1, y1) to (x2, y2) of the path segment to be detected, and the current robot index identifier; Enter the static detection phase, traverse the obstacle list, and perform the following sub-process for each obstacle: An axis-aligned bounding box fast detection algorithm is used to calculate the minimum Euclidean distance between the path segment and the obstacle as the shortest path distance; Perform static collision judgment. The static collision condition is whether the square of the shortest path distance is greater than the square of the sum of the obstacle radius and the robot body radius. If so, return False; otherwise, return True. Enter the dynamic detection phase, traverse the robot list, and perform the following sub-processes for each robot: When the traversed robot list index is equal to the current robot index identifier, the dynamic detection phase is skipped; Otherwise, determine the following dynamic collision conditions: Where path(t) is the linear interpolation position of the path segment, robot(t) is the predicted trajectory position of the robot; robot is the radius of the robot body; If the dynamic collision condition is not met, it returns False, otherwise it returns True; If True is returned, it means a collision is detected; If False is returned, it indicates that the path segment is safe, and the corresponding candidate path node passes the test and serves as the new extended node.

6. The system according to claim 1, wherein: After obtaining the new expansion node, an optimization radius is defined with the new expansion node as the center, all historical path nodes of the robot within the optimization radius are traversed, and an optimal path from the current position node to the new expansion node is generated as an alternative path; The robot's historical path node sequence and the parent relationship of all nodes in the robot's historical path node sequence are updated.

7. The system according to claim 1, wherein: After the robot completes a single path segment, it performs a status check, including but not limited to: calculating the Euclidean distance between the current position node and the target position node, If the Euclidean distance is less than the specified distance and the velocity vector converges, the driving task is marked as completed, the robot resources are released and the system status list is updated; otherwise, the driving task is marked as incomplete and the robot continues to enter the next driving task strategy execution cycle.

Images