A method and system for safe path planning of a mobile robot based on multiple constraints in a dynamic environment

Abstract

The application discloses a kind of dynamic environment-based mobile robot safety path planning method and system based on multiple constraints, the method is planned level by risk constraint modeling active elimination potential conflict between man and machine, and the planning framework of risk assessment based on cyclic A * multi-path algorithm generation is designed: first, the initial solution set of multiple paths is generated using cyclic A * algorithm combined with hard constraint conflict detection mechanism;Second, a path diversity filtering mechanism based on Jaccard similarity is designed to eliminate redundant paths to preserve solution set diversity;Finally, by fusing the random human risk probability model in the environment, the optimal safe path is selected from the candidate path through multi-dimensional risk calculation.The path planning method of the application solves the limitations of traditional algorithms in dynamic environments, which are difficult to predict human risk, and provides a safe and optimal control strategy and operation path for mobile robots through risk avoidance at the planning level.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, and in particular to a method and system for safe path planning of mobile robots based on multiple constraints in dynamic environments. Background Technology

[0002] As mobile robots are increasingly used in human-robot collaborative environments, ensuring their safety during task execution in dynamic human activity scenarios has become a critical challenge. In fields such as smart warehousing, a core issue when robots share workspaces with humans is that the randomness and unpredictability of human movement significantly increases the risk of collisions. While traditional path planning methods (such as the A* algorithm) can optimize efficiency, they do not fully consider the uncertainty of human behavior and cannot provide safe paths in dynamic environments. On the other hand, simple reinforcement learning algorithms suffer from poor training stability and large fluctuations in results in scenarios with multiple humans coexisting, making it difficult to significantly improve safety.

[0003] Existing research indicates that classic path planning algorithms (such as RRT and A*) can provide efficient target paths for mobile robots, but they do not fully consider the random risk factors of dynamic humans in the environment, resulting in a high risk of collisions. While reinforcement learning methods can adapt to dynamic environments, they suffer from insufficient training stability and significant fluctuations in task success rates in multi-person shared scenarios. Summary of the Invention

[0004] Purpose of the Invention: The purpose of this invention is to provide a method and system for safe path planning of mobile robots based on multiple constraints in dynamic environments. This addresses the problems of high conflict rates, poor path adaptability, and lack of safety constraints in existing mobile robot path planning technologies for human-robot shared environments, caused by the randomness of human behavior. The invention aims to accurately predict human movement trends through risk modeling, construct heterogeneous trajectory sets using a multi-path generator combined with active diversity control, and innovatively integrate hard constraint conflict resolution and soft constraint risk assessment mechanisms to collaboratively optimize the global path within a limited task time. Ultimately, this solves core defects such as robot obstacle avoidance failure and excessive task failure rates in highly dynamic scenarios, achieving a balance between significantly reduced conflict rates and a systematic improvement in task success rates, thus ensuring safety and performance.

[0005] Technical solution: A mobile robot safety path planning system based on multiple constraints in a dynamic environment, comprising:

[0006] The risk constraint construction module assesses the collision risks that may be caused by random human movement based on the acquired work environment and task information, constructs a human risk set, and builds soft and hard constraint rules.

[0007] The optimal safe path planning module is based on soft and hard constraint rules. Hard constraints are used to avoid high-risk areas, while soft constraints are responsible for path risk optimization to obtain the path with the minimum risk. This enables the planning of robot paths that are both safe and practical in complex dynamic environments.

[0008] Furthermore, the risk constraint construction module includes:

[0009] The stochastic simulation module is used to simulate the dynamic laws of uncertain factors, and follows the following principles: first, to ensure that there is no conflict between any individuals; second, to simulate the natural tendency of individuals to move towards the target direction.

[0010] The risk calculation module receives human movement trajectory data generated by the random simulation module. By statistically analyzing the probability that each grid location is occupied by humans at a specific time step, it accumulates the risk values ​​of all humans at the same spatiotemporal point to form a global risk assessment. Finally, it outputs a dynamically updated risk map that fully covers the comprehensive collision risk values ​​of all locations in the environment at different times.

[0011] The soft and hard constraint construction module transforms risk data into executable path planning rules. Hard constraints directly identify high-risk areas where robots are absolutely prohibited from entering, and these areas are forcibly marked as no-go zones to completely avoid collisions. Soft constraints establish a graded response mechanism for medium and low-risk areas. When the local risk is significantly higher than the average level of the environment, a path penalty mechanism is triggered to guide robots to detour but not completely prohibit passage.

[0012] Furthermore, the optimal safe path planning module includes:

[0013] The multi-path generation module automatically generates multiple alternative paths from the robot's starting point to the target point. It also evaluates and filters the generated path set, eliminating similar paths, and finally outputs a candidate path set containing multiple trajectories with different spatial distributions.

[0014] The conflict solver module ensures that the robot's planned path will not collide with the human during execution. It receives a set of candidate paths from the multi-path generation module and selects paths that satisfy all hard constraints to retain and pass to the risk level estimation module for processing.

[0015] The risk level estimation module quantifies the risk value faced by the candidate paths filtered by the conflict solver module when passing through each location, thereby calculating the risk level index representing the overall safety level of the path. By comparing the risk level indices of all remaining candidate paths, the path with the lowest comprehensive risk level is selected as the final optimal safe path output.

[0016] A multi-constraint-based safe path planning method for mobile robots in dynamic environments is proposed. Based on any of the aforementioned mobile robot safe path planning systems, it outputs the optimal safe path. The method assumes the robot operates in an environment where humans move randomly, and the robot needs to travel to a designated location to complete a specific task. The robot and... Individuals share the same environment, and each robot has a corresponding target location. Humans move within the environment according to a random strategy and are not controlled. The system is characterized by the following steps:

[0017] S1, the robot acquires information about the working environment and tasks;

[0018] S2, based on work environment and task information, constructs a human risk set using the risk constraint construction module; and constructs soft and hard constraints based on the average value of the human risk set.

[0019] S3, based on the optimal safe path planning module, constructs a planning framework that coordinates soft and hard constraints to obtain the path with the least risk;

[0020] S4, Determine if the robot's operation is interrupted: If the operation is not interrupted, return to step S1 to continue execution; if the operation is interrupted, the task ends.

[0021] Furthermore, the steps to construct a human risk set include:

[0022] S11, Computer Personnel The probability of choosing an action;

[0023] Personnel exist The state at time is Its target state is Given personnel status In this state, the personnel The set of actions is The robot steps at any time Take action , ;personnel In state Under the given conditions, prioritize actions that are closer in distance; the optimal action is... :

[0024] ,

[0025] in, Indicates the calculation personnel from To the target Manhattan distance;

[0026] personnel The probability of choosing an action is The expression is as follows:

[0027] ,

[0028] in, Represents all optimal actions The set, ; This indicates the number of actions in the corresponding action set; others do not belong to the optimal action set. Selection probability The sum of the probabilities of all actions at the current moment is 1.

[0029] S12, predicting personnel's state of observation through random simulation. The trajectory of the starting motion, the vertex Personnel The probability of occupying;

[0030] time step At that time, the vertex Personnel probability of occupying for: ,

[0031] in, This represents the total number of simulations. Indicates time vertex The number of times it is accessed; 0 < t ≤ T; Personnel at all times The state is represented as , representing a vertex Personnel occupy;

[0032] So, vertex At any moment By all Risk value shared by individuals Calculated using the following formula:

[0033] ,

[0034] S13, Obtain the human risk set :

[0035] ,

[0036] in, satisfy , , t0 ≤ t ≤ t0+T.

[0037] Furthermore, based on the average value of the human risk set, the principles for constructing soft and hard constraints are as follows:

[0038] when ,Will Defined as a hard constraint, meaning that in Time-based robots are prohibited from being occupied. Location;

[0039] when ,Will Defined as a soft constraint;

[0040] in, This represents the average value of the risk set. It is the scaling factor.

[0041] Furthermore, the specific implementation steps of the planning framework for coordinated soft and hard constraints include:

[0042] S31, through the multi-path generation module, uses the cyclic A* multi-path algorithm to construct multiple feasible paths from the starting point to the ending point, obtaining a multi-path set. ;

[0043] S32 uses a path diversity filtering algorithm to remove redundant paths from the initial path set and updates the multi-path set. ;

[0044] S33 assesses the uncertainty risk of human movement through the risk constraint construction module. The conflict solver module uses hard constraints to resolve the conflict between the robot and the human, and uses soft constraints to select the minimum risk path from the set of multiple paths, and then decides the minimum risk path from the selected path set.

[0045] Furthermore, in a given grid map ,starting point and the end point Under the given conditions, the steps for generating multiple paths using the cyclic A* multipath algorithm include:

[0046] S311, Initialize a multipath set to store the final result. and get the current time Personnel observation status ; Represents the empty set;

[0047] S312, Risk constraint set is generated by the risk constraint construction module. ,in, ,satisfy , , t0 ≤ t ≤ t0+T; ;

[0048] in, Represents the set of hard constraints, containing any time steps. Time risk value Greater than a given threshold vertex , ; Represents the set of soft constraints, containing any number of time steps. Time risk value Less than or equal to a given threshold vertex , ;

[0049] S313, Continue selecting the current evaluation function value smallest vertex Explore the area, and if the vertex is the target vertex, backtrack to build the path. and add to the set If the current search ends, the vertex is marked as explored, and all its neighboring vertices are checked. ;in It is accumulated from the starting point to the peak. The actual cost, yes To the target Heuristic cost estimation;

[0050] For each adjacent vertex Calculate its value from the current vertex Extended cumulative time step cost If this cumulative time step The total time step budget has been exceeded. If so, then skip that vertex;

[0051] If the vertex In time step Time risk value Then create the current vertex. A cloned vertex , The cost of extending the record for this cloned vertex. And calculate its evaluation value. Then, this cloned vertex is used to replace the original one. Further processing is required;

[0052] If the prediction is in In time step Time risk value And at the same time predict in In time step Time risk value Also skip that vertex;

[0053] For new vertices that have not been skipped, add them to the candidate set; for vertices already in the candidate set, if the newly calculated cumulative cost is... If better, then update the vertex pointer. Record new costs And recalculate ;

[0054] Each path obtained from a successful search ;

[0055] S314, Given two paths and Using functions Assess whether the two paths are the same. The expression is:

[0056] ,

[0057] in, It is the size of the intersection of the state sets of the two paths. It is the size of the union of the state sets of the two paths; diversity varies. from Increase to And enhance, Time indicates and They are the same path. This indicates that the two paths are completely different;

[0058] set up The path diversity threshold is defined for any two paths. The following constraints must be met:

[0059] ,

[0060] in, It is the generated set of paths.

[0061] Furthermore, the process of selecting the minimum risk path from the multi-path set based on soft constraints is as follows: In the generation of diverse paths, a dynamic set of hard risk constraints is used. High-risk vertices are masked, if the vertex At time step risk value Create virtual vertices Alternative Forced path detour; if adjacent vertices and At the same time belong to Skip the state transition, then proceed with soft constraint risk assessment and path risk quantification.

[0062] Furthermore, during the path optimization phase, the risk level estimation module receives the set of multipaths output by the multipath generator. By accumulating paths Calculate the risk level of each path by taking the risk values ​​of all vertices at the corresponding time step, and select the path with the lowest risk as the global optimal solution. Represents the candidate path set The Middle A path.

[0063] Compared with the prior art, the significant advantages of this invention are as follows:

[0064] 1. Significantly Optimized Conflict Rate: Compared with traditional algorithms, the Mobile Robot Global Path Planning (SFGPP) framework proposed in this invention significantly reduces human-robot conflict events in the robot path through the synergistic effect of hard-constraint conflict solvers and soft-constraint risk assessments. Experiments demonstrate that this method not only reduces conflicts in small maps but also maintains a low conflict level in large, high-density human-robot shared environments, with a more uniform conflict distribution, indicating its strong environmental adaptability.

[0065] 2. More Accurate Risk Modeling: The core of risk modeling is to construct dynamic spatiotemporal risks to quantify the probability of human-machine conflict. The risk constraint construction module of this invention predicts human movement trajectories through random simulation. Based on the goal-oriented principle, humans prioritize actions that shorten the distance to the target, and random movement behavior is simulated through conditional probability distribution. After multiple simulations to calculate the vertex occupancy probability, multiple human risks are aggregated to form a spatiotemporal risk set, which is intuitively represented by hard constraints and soft constraints. These are graded by thresholds and used for path node shielding and risk path optimization, respectively. The risk constraint construction module of this invention provides prior risk basis for path planning.

[0066] 3. More reliable path diversity assurance: This invention ensures the heterogeneity of the candidate path set through a cyclic A* algorithm and diversity filtering. It uses a backtracking mechanism to derive multiple suboptimal paths from the optimal path, and then uses Jaccard similarity to measure the spatial difference of the paths. A threshold is set to filter redundant paths: only paths with diversity higher than the threshold are retained, and redundant solutions with similar spatial distributions are removed. This design enables the final path set to cover a wider safety area, supporting the decision-making unit in screening low-risk paths. Attached Figure Description

[0067] Figure 1 This is a system architecture diagram of the present invention;

[0068] Figure 2 Construct a module diagram for risk constraints;

[0069] Figure 3 This is a diagram of the optimal safe path planning module;

[0070] Figure 4 This is a flowchart of the method of the present invention;

[0071] Figure 5 This is a schematic diagram of a simulated environment for human-robot collaboration in a warehouse setting.

[0072] Figure 6 This is a schematic diagram of the conflict distribution on a 10×20 map. Detailed Implementation

[0073] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0074] To address the path planning safety issues in human-robot shared environments, this invention proposes a safety-first global path planning framework for mobile robots (SFGPP). This framework employs a cyclic A*-based multi-path planning and optimization mechanism, comprising three stages: the first stage constructs multiple feasible paths from the starting point to the destination using a cyclic A* multi-path generator; the second stage uses a path diversity filtering algorithm to eliminate redundant paths from the initial path set; and the third stage assesses the uncertainty risk of human movement through a risk constraint construction mechanism and determines the safest path (i.e., the path with the minimum risk) from the filtered path set. Through stochastic risk modeling and multi-path optimization, the proposed global path planning framework proactively resolves potential random conflicts at the planning level, providing assurance for safe operation of mobile robots in uncertain environments.

[0075] like Figure 1 As shown, a mobile robot safety path planning system based on multiple constraints in a dynamic environment includes:

[0076] The core task of the risk constraint construction module is to assess the collision risk that may be caused by random human movement. In a two-dimensional mesh world experimental environment, this collision risk assessment is completed through stochastic simulation methods.

[0077] like Figure 2 As shown, the risk constraint construction module mainly consists of a stochastic simulation module, a risk calculation module, and a soft and hard constraint construction module. The stochastic simulation module is used to simulate the dynamic patterns of uncertain factors. To more realistically reflect natural human movement behavior, this stochastic simulation module follows two key principles: first, ensuring no conflict between any two individuals; and second, simulating the natural tendency of individuals to move towards a target direction. Based on these principles, this invention defines a human movement model.

[0078] The risk calculation module's core function is to transform the uncertainty of human behavior into quantifiable spatial safety risks. It receives human movement trajectory data generated by the stochastic simulation module, calculates the probability of each grid location being occupied by humans at a specific time step, and then accumulates the risk values ​​of all humans at the same spatiotemporal point to form a global risk assessment. This module ultimately outputs a dynamically updated risk map that comprehensively covers the combined collision risk values ​​of all locations in the environment at different times, providing a quantitative basis for subsequent planning.

[0079] The hard and soft constraint construction module transforms risk data into executable path planning rules. Hard constraints directly identify high-risk areas where robots are absolutely prohibited from entering—if the risk value of a location exceeds a preset threshold at a specific time, it is forcibly marked as a no-go zone to completely avoid collisions. Soft constraints establish a tiered response mechanism for low- to medium-risk areas; when the local risk is significantly higher than the environmental average, a path penalty mechanism is triggered, guiding robots to prioritize detours but not completely prohibiting passage. By integrating these two constraint strategies, core safety requirements are ensured while retaining flexible adjustment space for path optimization. The resulting composite constraint rules directly drive the optimal safe path planning module to generate safe and efficient motion paths.

[0080] The optimal safe path planning module is designed to plan robot paths that are both safe and practical in complex dynamic environments. This module employs an optimal safe path planning mechanism to obtain the path with the least risk. The overall workflow of this optimal safe path planning mechanism is as follows: Figure 3 As shown.

[0081] as follows Figure 3 As shown, this optimal safe path planning module mainly consists of a multi-path generation module based on a loop A*, based on...

[0082] The system comprises a conflict solver module based on hard constraints and a risk level estimation module based on soft constraints. The multi-path generation module, based on cyclic A*, automatically generates multiple candidate paths from the robot's starting point to the target point. It evaluates and filters the generated path set, eliminating similar paths and ultimately outputting a set of candidate paths with different spatially distributed trajectories. These candidate paths form the basis for subsequent conflict and risk assessments. The conflict solver module, based on hard constraints, ensures that the robot's planned path will not collide with the human during execution. It receives the candidate path set from the multi-path generation module and selects only paths that satisfy all hard constraints, passing them to the risk level estimation module for processing. The risk level estimation module, based on soft constraints, quantitatively evaluates the risk values ​​faced by the candidate paths filtered by the conflict solver module at various locations, calculating a risk level index representing the overall safety of the path. By comparing the risk level indices of all remaining candidate paths, it selects the path with the lowest overall risk level as the final optimal safe path.

[0083] A method for safe path planning of mobile robots based on multiple constraints in dynamic environments, the flowchart of which is shown below. Figure 4 As shown, the main steps are as follows:

[0084] Step 1: Set up the environment and task information;

[0085] This embodiment considers multiple robots operating in an environment where humans move randomly. The robots need to complete specific tasks at designated locations. The robots and... Individuals share the same environment, and each robot has a corresponding target location. Humans move within the environment according to random strategies and are uncontrolled, leading to uncertainty for the robots when accessing their target locations. Therefore, at any discrete time step... At that time, the robot may interact with a person. A conflict occurred.

[0086] The robot works on a two-dimensional grid ( ) world, in which , Let be a positive integer, defining the size of the 2D grid world environment, which can be represented as: .in Represents vertices The set, Represents the set of edges. It is the set of robot states. ,in and Represents vertices Coordinates in a grid world; This is the robot's initial state. This is the robot's target state.

[0087] Step 2: Construct a risk set based on the risk constraint construction module;

[0088] In a two-dimensional grid environment, by quantifying the dynamic risks posed by humans, spatiotemporal constraints are provided for the safe path planning of robots. exist The state at time is Its target state is Given personnel status In this state, the personnel The set of actions is The robot takes a step at any given time. Take action ( ),personnel In state Under the given conditions, select an action The probability is Humans prioritize actions that are closer in distance; their optimal action is... :

[0089] (1)

[0090] in Computing personnel from To the target Manhattan distance.

[0091] The conditional probability distribution expression for human action choices is as follows:

[0092] (2)

[0093] in, Indicates the set of priority actions. , which includes all actions that satisfy equation (1); priority actions are given higher conditional probabilities to prompt people to move in a goal-oriented manner. This indicates the number of actions in the corresponding action set. ; , indicating that the sum of the probabilities of all actions is 1.

[0094] Risk calculation: In a two-dimensional grid environment, observations were made. Personnel at all times The state is represented as , where the vertex Personnel Occupation. In this embodiment, random simulation is used to predict the human's state of observation. The starting trajectory allows us to calculate the following conditional probabilities: at a time step... (0 < t ≤ T), vertex Personnel probability of occupying for: (3)

[0095] in, This represents the total number of simulations. Indicates time vertex Number of times it was accessed.

[0096] So, vertex At any moment By all Risk value shared by individuals It can be calculated using the following formula:

[0097] (4)

[0098] Ultimately, a set of human risks is obtained. :

[0099] (5)

[0100] in, satisfy: , , t0 ≤ t ≤ t0+T.

[0101] Construction of soft and hard constraints: express The average value of the risk set at time t. It is the scaling factor. ,when ,exist Robots must not be allowed to occupy the space at all times. The location of all risk values ​​is called a hard constraint. > Location All are recorded in the hard constraint set Otherwise, it is a soft constraint and is recorded in the soft constraint set. ; .

[0102] The specific algorithms for human stochastic risk assessment and the construction of soft and hard constraints are shown in Table 1. Based on the risk constraint construction module, the human stochastic risk assessment algorithm is used to obtain the human risk set. .

[0103] Table 1. Algorithms for Human Random Risk Assessment and Construction of Soft and Hard Constraints

[0104]

[0105] Step 3: Based on the optimal safe path planning module, plan the optimal safe path;

[0106] Step 31: Generate multiple paths based on the cyclic A* multipath algorithm;

[0107] The Looping A* multipath algorithm is used in a given grid map. ,starting point and the end point This algorithm works on the basis of [previous algorithm]. It first initializes a multipath set to store the final result. and get the current time Personnel observation status “ "" indicates an empty set. The risk constraint set is generated by the risk constraint construction module. ,for , ;in, for The set of hard constraints at time t represents time t. High-risk collision peak The set, ; for The set of soft constraints at time t represents the time t. The risk value is greater than And less than the risk mean Vertices (as a given threshold) gather, .generally However, it can also be given directly based on the actual situation.

[0108] The Looping A* multipath algorithm maintains a candidate set of nodes to be explored and a record of already explored nodes in its main loop. It continuously selects the current evaluation function value. smallest node (i.e., vertex) ) to explore, among which It is accumulated from the starting point to the node. The actual cost, yes To the target Heuristic cost estimation. Once the node is the target node, the construction path is backtracked. And add to the multi-path set If the current search ends, the node is marked as explored, and all its neighboring nodes are checked. For each adjacent node The algorithm calculates its value from the current node. Extended cumulative time step cost , It is usually set to a step size, that is =1. If this cumulative time step cost The total time step budget has been exceeded. If so, then skip that node.

[0109] The next crucial step is risk aversion: if the prediction is in... In the exploratory phase In the set of hard constraints, If the node is not found, skip it and explore other nodes. Additionally, check for edge conflicts; if... and That is, from Time to time High risk is caused by Passed to At this time, the robot is Transferred to If the condition is met, it will be considered an edge conflict, and the exploration of other nodes will be skipped. Other cases will be handled as follows:

[0110] Create the current node A clone node Record the expansion cost for this clone node. And calculate its evaluation value. Then, this cloned node replaces the original one. Further processing is performed. The new node is added to the candidate set; for nodes already in the candidate set, if the newly calculated cumulative cost... If better, then update the node pointer to point to. Record new costs And recalculate .

[0111] The multipath set is obtained by using the loop A* multipath algorithm. .

[0112] For each path obtained from a successful search It also needs to undergo path diversity filtering: given two paths and Using functions Assess the diversity of both and obtain a diversity score. . The Jaccard similarity calculation expression is as follows:

[0113] (6)

[0114] in, It is the size of the intersection of the state sets of the two paths. It is the size of the union of the state sets of the two paths. Diversity varies. from Increase to And enhance, Time indicates and They are the same path. This indicates that the two paths are completely different.

[0115] The goal of this embodiment is to generate multiple paths while maintaining their diversity. Let... The threshold for path diversity. For any two paths The following constraints must be met:

[0116] (7)

[0117] The algorithm only retains those that are stored along with all others. Diversity score between paths All are not lower than the preset threshold. The path is chosen to satisfy the diversity constraint. Ultimately, the output of the diversity path satisfies the risk constraint. and the aforementioned diversity constraints Set of multiple accessible paths .

[0118] Step 32: Based on the hard constraint conflict solver, high-risk nodes are shielded;

[0119] First, a hard-constraint conflict solver is needed. To quantify the risks of human-computer interaction, two types of conflicts need to be clearly defined: vertex conflicts and edge conflicts. These conflicts constitute the hard constraints of path planning. Vertex conflicts occur when the robot is in a certain state... With the human state Occupying the same vertex at the same time, that is The path is invalid when the robot and the human exchange positions in adjacent time steps; an edge conflict occurs when the robot and the human exchange positions in adjacent time steps. The path is prohibited.

[0120] Conflict resolution mechanism: In the generation of diverse paths, through a set of hard constraints Block high-risk nodes, if nodes At time step risk value Forced path detour; if adjacent nodes and If there is a potential edge conflict, skip the state transition, followed by soft constraint risk assessment and path risk quantification.

[0121] Step 33: Based on the risk level estimation of soft constraints, a quantitative assessment of dynamic risks in the environment is achieved;

[0122] Soft constraint risk level estimation quantifies dynamic risks in the environment by establishing stochastic models of human behavior, providing a safety basis for core path decisions. Its operating principle begins with the current moment... Observed human condition Multiple random trajectory simulations were performed (using the human random risk assessment algorithm shown in Table 1) to calculate the future trajectory. Any vertex within a step At time step By humans Conditional probability of occupancy ,in This represents the total number of simulations. This indicates the number of times the vertex was visited in the simulation. (Combining all...) After considering the risk impact on celebrities, define space-time nodes. Risk value This ultimately forms a time-progressing set of risks. .

[0123] During the path optimization phase, the risk level estimation module receives the set of multipaths output by the multipath generator. By accumulating paths The risk level of each path is calculated based on the risk values ​​of all vertices at the corresponding time step, and the path with the lowest risk is selected as the global optimal solution. This process not only avoids conflicts caused by hard constraints, but also achieves path safety optimization through probabilistic risk assessment. Combined with the diverse constraints imposed in the multi-path generation stage, it constitutes a complete planning framework that coordinates soft and hard constraints.

[0124] Step 4: Generating the minimum risk path;

[0125] For candidate paths Comprehensive risks Calculated by dynamic weighted average of spatiotemporal risks: ,in For path Length, From formula (4), we can obtain the result. Based on formula (8), select the path with the minimum risk:

[0126] (8)

[0127] in, Represents the candidate path set The Middle A path.

[0128] Step 5: Determine if the robot's operation has been interrupted;

[0129] If the operation is not interrupted, return to step 1 to continue execution;

[0130] If the task is interrupted, the task ends.

[0131] like Figure 5 The image shown is a simulation environment for this embodiment, with a map size of 10×20. The grid includes static obstacles (black squares) and the robot's starting point. Target point The initial positions of the two humans , Target location , .

[0132] In problem modeling, the parameters are set as follows: time step budget. Number of risk simulations Path diversity threshold The reward function of the proposed cyclic A* multipath algorithm is defined in Table 2.

[0133] Table 2. Parameters of the Reward Function

[0134]

[0135] Figure 6 The conflict distribution is shown on a 10×20 map. Multiple methods were used in the experiment to simulate and plan safe paths multiple times, validating the proposed cyclical approach from multiple dimensions. The effectiveness of the algorithm is shown in Table 3.

[0136] Table 3. Comparison of average conflict rate and success rate for each method

[0137]

[0138] This embodiment demonstrates that in a 10×20 grid environment, through human risk modeling, multi-path generation, and diversity constraint screening, the proposed framework reduces the conflict rate to 0.12. Experimental results show that the cyclic A* multi-path algorithm of this invention can reduce conflicts, improve the success rate, and support safe global path planning for mobile robots in a shared environment.

Claims

1. A mobile robot safety path planning system based on multiple constraints in a dynamic environment, characterized in that, include: The risk constraint construction module assesses the collision risks that may be caused by random human movement based on the acquired work environment and task information, constructs a human risk set, and builds soft and hard constraint rules. The optimal safe path planning module is based on soft and hard constraint rules. Hard constraints are used to avoid high-risk areas, while soft constraints are responsible for path risk optimization to obtain the path with the minimum risk. To enable the planning of robot paths that are both safe and practical in complex and dynamic environments.

2. The mobile robot safety path planning system based on multiple constraints in a dynamic environment according to claim 1, characterized in that, The risk constraint construction module includes: The stochastic simulation module is used to simulate the dynamic laws of uncertain factors, and follows the following principles: first, to ensure that there is no conflict between any individuals; second, to simulate the natural tendency of individuals to move towards the target direction. The risk calculation module receives human movement trajectory data generated by the random simulation module. By statistically analyzing the probability that each grid location is occupied by humans at a specific time step, it accumulates the risk values ​​of all humans at the same spatiotemporal point to form a global risk assessment. Finally, it outputs a dynamically updated risk map that fully covers the comprehensive collision risk values ​​of all locations in the environment at different times. The soft and hard constraint construction module transforms risk data into executable path planning rules. Hard constraints directly identify high-risk areas where robots are absolutely prohibited from entering, and these areas are forcibly marked as no-go zones to completely avoid collisions. Soft constraints establish a graded response mechanism for medium and low-risk areas. When the local risk is significantly higher than the average level of the environment, a path penalty mechanism is triggered to guide robots to detour but not completely prohibit passage.

3. The mobile robot safety path planning system based on multiple constraints in a dynamic environment according to claim 1, characterized in that, The optimal safe path planning module includes: The multi-path generation module automatically generates multiple alternative paths from the robot's starting point to the target point. It also evaluates and filters the generated path set, eliminating similar paths, and finally outputs a candidate path set containing multiple trajectories with different spatial distributions. The conflict solver module ensures that the robot's planned path will not collide with the human during execution. It receives a set of candidate paths from the multi-path generation module and selects paths that satisfy all hard constraints to retain and pass to the risk level estimation module for processing. The risk level estimation module quantifies the risk value faced by the candidate paths filtered by the conflict solver module when passing through each location, thereby calculating the risk level index representing the overall safety level of the path. By comparing the risk level indices of all remaining candidate paths, the path with the lowest comprehensive risk level is selected as the final optimal safe path output.

4. A method for safe path planning of a mobile robot in a dynamic environment based on multiple constraints, based on the mobile robot safe path planning system as described in any one of claims 1-3, to output the optimal safe path; setting the robot to operate in an environment where humans move randomly, and the robot needs to go to a designated location to complete a specific task; wherein the robot and... Individuals share the same environment, and each robot has a corresponding target location. Humans move within the environment according to a random strategy and are not subject to control. Its characteristic is that... The steps include the following: S1, the robot acquires information about the working environment and tasks; S2, based on work environment and task information, constructs a set of human risks using a risk constraint construction module; Construct soft and hard constraints based on the average value of the human risk set; S3, based on the optimal safe path planning module, constructs a planning framework that coordinates soft and hard constraints to obtain the path with the least risk; S4, Determine if the robot operation is interrupted: If the operation is not interrupted, return to step S1 to continue execution; If the task is interrupted, the task ends.

5. The method for safe path planning of mobile robots based on multiple constraints in dynamic environments according to claim 4, characterized in that, The steps to construct a human risk set include: S11, Computer Personnel The probability of choosing an action; Personnel exist The state at time is Its target state is Given personnel status In this state, the personnel The set of actions is The robot steps at any time Take action as , ;personnel In state Under the given conditions, prioritize actions that are closer in distance; the optimal action is... : ， in, Indicates the calculation personnel from To the target Manhattan distance; personnel The probability of choosing an action is The expression is as follows: ， in, Represents all optimal actions The set, ; This indicates the number of actions in the corresponding action set; others do not belong to the optimal action set. Selection probability The sum of the probabilities of all actions at the current moment is 1. S12, predicting personnel's state of observation through random simulation. The trajectory of the starting motion, the vertex Personnel The probability of occupying; time step At that time, the vertex Personnel probability of occupying for: , in, This represents the total number of simulations. Indicates time vertex The number of times it is accessed; 0 < t ≤ T; Personnel at all times The state is represented as , representing a vertex Personnel occupy; So, vertex At any moment By all Risk value shared by individuals Calculated using the following formula: ， S13, Obtain the human risk set : ， in, satisfy , , t0 ≤ t ≤ t0+T.

6. The method for safe path planning of mobile robots based on multiple constraints in dynamic environments according to claim 5, characterized in that, Based on the average value of the human risk set, the principles for constructing soft and hard constraints are as follows: when ,Will Defined as a hard constraint, meaning that in Time-based robots are prohibited from being occupied. Location; when ,Will Defined as a soft constraint; in, This represents the average value of the risk set. It is the scaling factor.

7. The method for safe path planning of mobile robots based on multiple constraints in dynamic environments according to claim 4, characterized in that, The specific implementation steps of the planning framework that coordinates soft and hard constraints include: S31, through the multi-path generation module, uses the cyclic A* multi-path algorithm to construct multiple feasible paths from the starting point to the ending point, obtaining a multi-path set. ; S32 uses a path diversity filtering algorithm to remove redundant paths from the initial path set and updates the multi-path set. ; S33 assesses the uncertainty risk of human movement through the risk constraint construction module. The conflict solver module uses hard constraints to resolve the conflict between the robot and the human, and uses soft constraints to select the minimum risk path from the set of multiple paths, and then decides the minimum risk path from the selected path set.

8. The method for safe path planning of mobile robots based on multiple constraints in a dynamic environment according to claim 7, characterized in that, In a given grid map ,starting point and the end point Under the given conditions, the steps for generating multiple paths using the cyclic A* multipath algorithm include: S311, Initialize a multipath set to store the final result. and get the current time Personnel observation status ; Indicates the empty set; S312, Risk constraint set is generated by the risk constraint construction module. ,in, ,satisfy , , t0 ≤ t ≤ t0+T; ; in, Represents the set of hard constraints, containing any time steps. Time risk value Greater than a given threshold vertex , ; Represents the set of soft constraints, containing any number of time steps. Time risk value Less than or equal to a given threshold vertex , ; S313, Continue selecting the current evaluation function value smallest vertex Explore the area, and if the vertex is the target vertex, backtrack to build the path. and add to the set If the current search ends, the vertex is marked as explored, and all its neighboring vertices are checked. ;in It is accumulated from the starting point to the peak. The actual cost, yes To the target Heuristic cost estimation; For each adjacent vertex Calculate its value from the current vertex Extended cumulative time step cost If this cumulative time step The total time step budget has been exceeded. If so, then skip that vertex; If the vertex In time step Time risk value Then create the current vertex. A cloned vertex , The cost of extending the record for this cloned vertex. And calculate its evaluation value. Then, this cloned vertex is used to replace the original one. Further processing is required; If the prediction is in In time step Time risk value And at the same time predict in In time step Time risk value Also skip that vertex; For new vertices that have not been skipped, add them to the candidate set; for vertices already in the candidate set, if the newly calculated cumulative cost is... If better, then update the vertex pointer. Record new costs And recalculate ; Each path obtained from a successful search ; S314, Given two paths and Using functions Assess whether the two paths are the same. The expression is: ， in, It is the size of the intersection of the state sets of the two paths. It is the size of the union of the state sets of the two paths; diversity varies. from Increase to And enhance, Time indicates and They are the same path. This indicates that the two paths are completely different; set up The path diversity threshold is defined for any two paths. The following constraints must be met: ， in, It is the generated set of paths.

9. The method for safe path planning of mobile robots based on multiple constraints in a dynamic environment according to claim 7, characterized in that, The process of selecting the minimum risk path from a multi-path set based on soft constraints is as follows: In the generation of diverse paths, a dynamic set of hard risk constraints is used. High-risk vertices are masked, if the vertex At time step Risk value Create virtual vertices Alternative Forced path detour; if adjacent vertices and At the same time belong to Skip the state transition, then proceed with soft constraint risk assessment and path risk quantification.

10. The method for safe path planning of mobile robots based on multiple constraints in a dynamic environment according to claim 7, characterized in that, During the path optimization phase, the risk level estimation module receives the set of multipaths output by the multipath generator. By accumulating paths Calculate the risk level of each path by taking the risk values ​​of all vertices at the corresponding time step, and select the path with the lowest risk as the global optimal solution. Represents the candidate path set The Middle A path.

Images