Robot real-time path planning system and method based on predictive trigger mechanism

Abstract

The invention discloses a robot real-time path planning system and method based on a prediction trigger mechanism, and belongs to the field of mobile robos.The robot real-time path planning system comprises a map processing module, a path construction module, a dynamic obstacle sensing module, a collision risk prediction module, a path repairing module and a path updating module which are connected with one another; path planning is completed through the steps of sample point generation and path tree initialization, bidirectional path tree construction, path execution and root node updating, dynamic obstacle perception and trajectory prediction, collision risk detection and local path repair, end point detection and task completion. According to the method, the autonomous navigation performance of the mobile robot in a complex environment can be remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of mobile robots, and in particular to a robot real-time path planning system and method based on a prediction trigger mechanism. Background Art

[0002] With the development of artificial intelligence and intelligent manufacturing, mobile robots are playing an increasingly important role in industrial production, logistics and transportation, urban cleaning, and medical delivery. In these scenarios, robots often need to navigate complex environments with both static and dynamic obstacles. Therefore, achieving efficient, reliable, and real-time path planning has become a major research and engineering challenge in this field.

[0003] Traditional path planning methods primarily include graph search-based algorithms (such as Dijkstra, A*, and D) and sampling-based algorithms (such as RRT, RRT*, and FMT*). While graph search algorithms offer advantages in terms of path shortestness and completeness, they suffer from high computational complexity in high-dimensional spaces or large-scale environments, making them unsuitable for real-time planning. Sampling algorithms, such as RRT, offer good scalability, but their search efficiency decreases in obstacle-dense environments, and the generated paths often exhibit incoherence and excessive detours.

[0004] Especially in dynamic environments, traditional path planning algorithms require frequent path reconstruction or resampling, resulting in high computational resource consumption, delayed robot response, and even failure. Therefore, there is an urgent need for an efficient motion planning solution that combines global and local path planning and offers both real-time and predictive capabilities. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a robot real-time path planning system and method based on a predictive trigger mechanism (i.e., algorithm P-RT-BFMT), which aims to solve the key technical problems faced by mobile robots in performing efficient path planning in dynamic environments, such as poor real-time performance, insufficient obstacle avoidance capability, and unstable path. By introducing a series of innovative module designs and path reconstruction strategies, the autonomous navigation performance of mobile robots in complex environments is significantly improved.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: a robot real-time path planning system based on a prediction trigger mechanism, characterized in that it includes two groups of parallel modules, one group is a map processing module and a path construction module connected in sequence, and the other group is a dynamic obstacle perception module, a collision risk prediction module and a path repair module connected in sequence, and the path construction module and the path repair module are simultaneously connected to the path update module.

[0007] The further improvement of the technical solution of the present invention is that the specific structure and function of each module are as follows:

[0008] Map processing module: responsible for loading the map data of the robot's working environment and preprocessing it, i.e. grayscale Figure 2 The static obstacle space set X is finally extracted by performing value and rasterization operations. Fobs With Free Space Collection X free ;

[0009] Path construction module: Take the starting point and the target point as the root and construct the forward tree T f and the backward tree T r ,Using the improved bidirectional FMT* algorithm to perform path expansion and connection,,speeding up the path search process;

[0010] Dynamic obstacle perception module: continuously obtains the position and speed information of dynamic obstacles around the robot and records them in real time to support subsequent trajectory prediction processing;

[0011] Collision risk prediction module: This module estimates the future movement trajectory of dynamic obstacles based on their historical positions and velocities, and determines whether they may collide with the robot's current path.

[0012] Path Repair Module: When the system predicts an impending collision, it triggers local path reconstruction and re-searches and repairs only the blocked path locations, improving the algorithm's local responsiveness and computational efficiency.

[0013] Path update module: controls the robot to move along the currently planned path, and updates the root node, subpath, and execution cost in real time during the process to ensure the consistency and optimality of navigation execution.

[0014] A real-time path planning method for a robot based on a predictive trigger mechanism, the specific steps are as follows:

[0015] Step 1: Map loading and processing: The system loads the raster map or other environmental mapping data, obtains the static obstacle space information and free space area through binarization processing, and forms a basic navigation environment model;

[0016] Step 2: Sample point generation and path tree initialization: Use the SampleFree algorithm to randomly and uniformly sample N collision-free nodes in free space to construct a spatial sampling graph for path planning, reducing the amount of repeated sampling and collision detection calculations during runtime;

[0017] Step 3: Bidirectional path tree construction: The starting point and the end point are used as the root nodes of the forward tree and the backward tree, respectively. The bidirectional fast marching tree (BFMT) expansion mechanism is used to expand the path in each direction. When the distance between the nodes of the two trees is less than the set connection threshold, the path connection and merging operations are performed to generate the initial feasible path.

[0018] Step 4: Path execution and root node update: The robot moves along the planned path step by step. Whenever the robot reaches a key node on the path, it automatically sets the node as the new path tree root node and updates the path structure to ensure path coherence and cost optimality.

[0019] Step 5: Dynamic obstacle perception and trajectory prediction: When a moving obstacle enters the robot's perception range, the system initiates motion trajectory tracking and linear velocity estimation to predict and analyze the obstacle's future position.

[0020] Step 6: Collision risk detection and local path repair: The system compares the predicted obstacle trajectory with the current robot's expected path in time and space. If there is a potential collision risk, the local path reconstruction mechanism is triggered to regenerate a collision-free feasible path in the affected area and replace the original path segment.

[0021] Step 7: End point detection and task completion: When the robot enters the neighborhood of the end point, the system determines that the navigation task is completed and ends the path tracking execution.

[0022] The further improvement of the technical solution of the present invention is that the specific steps of step 1 are as follows:

[0023] Step 1.1: Load map information from SLAM or predefined CAD drawings;

[0024] Step 1.2: Perform a binarization operation on the map to identify the static obstacle area X Fobs Free movement area X free ;

[0025] Step 1.3: Divide the free space into discrete computing units to provide basic data structure support for path search and collision detection.

[0026] The further improvement of the technical solution of the present invention is that the specific steps of step 2 are as follows:

[0027] Step 2.1: Call the SampleFree(N) algorithm, at X free Generate N collision-free random sample points;

[0028] Step 2.2: Set the starting point x s is the forward tree T f The root node, target point x g is the backward tree T r The root node of

[0029] Step 2.3: Initialize the open node set:

[0030] V f-open ={x s},V r-open ={xg}.

[0031] The further improvement of the technical solution of the present invention is that the specific steps of step 3 are as follows:

[0032] Step 3.1: The system expands the forward tree T alternately or in parallel in each iteration f and the backward tree T r ;

[0033] Step 3.2: Use the following heuristic cost function to guide the node expansion direction:

[0034] Heuristic(x)=c(x)+‖xx g ‖

[0035] When the Euclidean distance between any two nodes in the two trees is less than the threshold δ, the path connection operation TreeMerge is performed.

[0036] A further improvement of the technical solution of the present invention is that the specific steps of step 4 are as follows:

[0037] Step 4.1: The robot moves along the currently connected path segment by segment. When the robot reaches the path node x k , set it as the new root node;

[0038] Step 4.2: Trigger the path cost recalculation mechanism and update the path costs of its descendant child nodes:

[0039] c(x q )=c(x p )+‖x q -x p ‖,x p =parent(x q )

[0040] Among them, c(x p ) refers to the current node cost, ‖x q -x p ‖ refers to the Euclidean distance between two nodes, x p is x q The parent node of .

[0041] The further improvement of the technical solution of the present invention is that: Step 5 is as follows: every interval of Δt seconds, obtain the position information L of the dynamic obstacle at time t1 and time t2 t1 ,L t2 , estimate its speed:

[0042]

[0043] Predict its future position sequence within the T=2s time window:

[0044]

[0045] At the same time, calculate the predicted trajectory Y of the robot within the path segment robot (t), if the following conditions are met:

[0046] ‖Y robot (t)-Y Dobs (t)‖ <R s

[0047] This indicates that there is a potential conflict in the path, and the system triggers the local path reconstruction mechanism.

[0048] The further improvement of the technical solution of the present invention is that: Step 6 specifically includes the following steps: adding all blocked path nodes to the queue to be repaired M O ;

[0049] Re-execute the Near neighbor node selection and Collisionless collision check operations in the local area, find the replacement parent node, build a new sub-path segment and replace the original path to complete the local reconstruction.

[0050] The further improvement of the technical solution of the present invention is that: Step 7 specifically comprises the following steps: determining whether the current position of the robot satisfies:

[0051] ‖x robot -x g ‖<=r g

[0052] If the path has entered the neighborhood of the end point, the task is considered completed; otherwise, the path execution and prediction operations are repeated.

[0053] Due to the adoption of the above technical solution, the present invention has achieved the following technical advancements: High-efficiency path planning capability: Through the BFMT bidirectional expansion mechanism, the path construction time is significantly shortened. In experimental comparison, the path generation time is reduced by about 40% compared with the RT-RRT* algorithm; the local path reconstruction frequency is reduced, which reduces the robot control and computing burden;

[0054] Robust dynamic obstacle avoidance: The prediction mechanism combined with the local reconstruction strategy can effectively deal with unexpected obstacle insertions and dynamically maintain path stability and continuity;

[0055] Path execution consistency and low cost: Root node updates are combined with path reconstruction to ensure that the robot's path does not fluctuate during movement, resulting in optimized execution costs.

[0056] High success rate and strong deployability: In various simulation environments, such as Sample, Narrow, and Maze, the planned paths are smooth, low-cost, and have good real-time performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention, and those skilled in the art can derive other drawings based on these drawings without inventive effort.

[0058] Figure 1 Schematic diagram of the system structure of the present invention;

[0059] Figure 2 It is a schematic flow chart of the method of the present invention;

[0060] Figure 3 It is a schematic diagram of bidirectional expansion connection;

[0061] Figure 4 It is a schematic diagram of the dynamic obstacle prediction and triggering mechanism;

[0062] Figure 5 It is a flowchart of local path repair and splicing;

[0063] Figure 6 There are three simulation environment maps: Sample (simple environment), Narrow (narrow passage) and Maze (maze structure);

[0064] Figure 7 It is a schematic diagram of dynamic obstacle avoidance in three simulation environments;

[0065] Figure 8 This is a comparison chart of average execution cost and arrival time under three simulation environments;

[0066] Figure 9 is the average success rate of different algorithms in three simulation environments. DETAILED DESCRIPTION

[0067] In order to more fully understand the technical solution of the present invention, the specific implementation of the robot real-time path planning system and method based on the prediction trigger mechanism provided by the present invention is now described in detail with reference to the accompanying drawings. Figures 1 to 8 Provide detailed explanation of the structure, algorithm flow, mechanism design and experimental verification.

[0068] System composition structure

[0069] like Figure 1Figure 2 shows the structure of the proposed system. The system primarily consists of the following six functional modules: two parallel groups of modules: a map processing module and a path building module, and a dynamic obstacle perception module, a collision risk prediction module, and a path repair module. The path building module and the path repair module are both connected to the path update module. These modules work together to complete path planning and obstacle avoidance navigation in dynamic environments:

[0070] Map processing module: responsible for loading the map data of the robot's working environment and preprocessing it, i.e. grayscale Figure 2 The static obstacle space set X is finally extracted by performing value and rasterization operations. Fobs With Free Space Collection X free ;

[0071] Path construction module: Take the starting point and the target point as the root and construct the forward tree T f and the backward tree T r , bidirectional FMT* algorithm is used for path expansion and connection to accelerate the path search process;

[0072] Dynamic obstacle perception module: continuously obtains the position and speed information of dynamic obstacles around the robot and records them in real time to support subsequent trajectory prediction processing;

[0073] Collision risk prediction module: This module estimates the future movement trajectory of dynamic obstacles based on their historical positions and velocities, and determines whether they may collide with the robot's current path.

[0074] Path Repair Module: When the system predicts an impending collision, it triggers local path reconstruction and re-searches and repairs only the blocked path locations, improving the algorithm's local responsiveness and computational efficiency.

[0075] Path update module: controls the robot to move along the currently planned path, and updates the root node, subpath, and execution cost in real time during the process to ensure the consistency and optimality of navigation execution.

[0076] Path planning method process

[0077] like Figure 2 As shown in FIG, the overall process of the path planning method of the present invention is demonstrated, including key steps such as map loading, sample generation, bidirectional expansion, path connection, navigation execution, collision prediction and local reconstruction.

[0078] The specific steps are as follows:

[0079] Step 1: Map loading and processing: The system loads the raster map or other environmental mapping data, obtains the static obstacle space information and free space area through binarization processing, and forms a basic navigation environment model;

[0080] Step 1.1: Load map information from SLAM or predefined CAD drawings;

[0081] Step 1.2: Perform a binarization operation on the map to identify the static obstacle area X Fobs Free movement area X free ;

[0082] Step 1.3: Divide the free space into discrete computing units to provide basic data structure support for path search and collision detection.

[0083] Step 2: Sample point generation and path tree initialization: Use the SampleFree algorithm to randomly and uniformly sample N collision-free nodes in free space to construct a spatial sampling graph for path planning, reducing the amount of repeated sampling and collision detection calculations during runtime;

[0084] Step 2.1: Call the SampleFree(N) algorithm, at X free Generate N collision-free random sample points;

[0085] Step 2.2: Set the starting point x s is the forward tree T f The root node, target point x g is the backward tree T r The root node of

[0086] Step 2.3: Initialize the open node set: All sampled nodes are divided into open set, closed set and unvisited set, laying the foundation for the subsequent heuristic expansion and state maintenance mechanism.

[0087] V f-open ={x s},V r-open ={x g}.

[0088] Step 3: Bidirectional path tree construction: Figure 3 As shown in the figure, the bidirectional expansion process under the BFMT structure is demonstrated. The starting point and the end point are used as the root nodes of the forward tree and the backward tree respectively. The bidirectional fast marching tree (BFMT) expansion mechanism is used to expand the path in each direction. When the distance between the nodes of the two trees is less than the set connection threshold, the path connection and merging operations are performed to generate the initial feasible path.

[0089] Step 3.1: The system expands the forward tree T alternately or in parallel in each iteration f and the backward tree Tr ;

[0090] Step 3.2: Node expansion adopts a heuristic strategy, giving priority to nodes with smaller total cost for neighbor connection. The cost function is as follows:

[0091] Heuristic(x)=c(x)+‖xx g ‖

[0092] When the Euclidean distance between any two nodes in the two trees is less than the threshold δ, the path connection operation TreeMerge is performed.

[0093] Step 4: Path execution and root node update: The robot moves along the planned path step by step. Whenever the robot reaches a key node on the path, it automatically sets the node as the new path tree root node and updates the path structure to ensure path coherence and cost optimality.

[0094] Step 4.1: The robot moves along the currently connected path segment by segment. When the robot reaches the path node x k , set it as the new root node;

[0095] Step 4.2: Trigger the path cost recalculation mechanism and update the path costs of its descendant child nodes:

[0096] c(x q )=c(x p )+‖x q -x p ‖,x p =parent(x q )

[0097] Among them, c(x p ) refers to the current node cost, ‖x q -x p ‖ refers to the Euclidean distance between two nodes, x p is x q This "root node update mechanism" ensures that the path structure remains coherent and optimal under local disturbances, supports event-driven replanning triggering logic, and collaborates with the dynamic perception module to achieve navigation optimization.

[0098] Step 5: Dynamic obstacle perception and trajectory prediction: Figure 4 The figure shows the principle of the dynamic obstacle detection and prediction mechanism of the present invention. When a moving obstacle enters the robot's sensing range, the system starts motion trajectory tracking and linear velocity estimation to predict the obstacle's future position.

[0099] Every Δt seconds, the position information L of the dynamic obstacle at time t1 and time t2 is obtained. t1 ,Lt2 , estimate its speed:

[0100]

[0101] Predict its future position sequence within the T=2s time window:

[0102]

[0103] At the same time, calculate the predicted trajectory Y of the robot within the path segment robot (t), if the following conditions are met:

[0104] ‖Y robot (t)-Y Dobs (t)‖ <R s

[0105] This indicates that there is a potential conflict in the path, and the system triggers the local path reconstruction mechanism.

[0106] Step 6: Collision risk detection and local path repair: The system compares the predicted obstacle trajectory with the current robot's expected path in time and space. If there is a potential collision risk, the local path reconstruction mechanism is triggered to regenerate a collision-free feasible path in the affected area and replace the original path segment.

[0107] Figure 5 The local path repair and splicing process is described. When the path is predicted to be blocked, the blocked node is added to the reconstruction queue, and a collision-free connection node is searched again in its neighborhood; all blocked path nodes are added to the queue to be repaired M. O The Near neighbor node selection and Collisionless collision check operations are re-executed in the local area, and the new parent node is found and the child node cost is updated in turn. After the connection is successfully established, the new path segment is spliced ​​into the main path and the original blocked path segment is replaced. During the whole process, the system dynamically updates the node status set (open, closed, unvisited, blocked, etc.) to maintain the consistency of the tree structure and the monotonicity of the cost.

[0108] Step 7: End point detection and task completion: When the robot enters the neighborhood of the end point, the system determines that the navigation task is completed and ends the path tracking execution.

[0109] Determine whether the robot's current position satisfies:

[0110] ‖x robot -x g ‖<=r g

[0111] If the path has entered the neighborhood of the end point, the task is considered completed; otherwise, the path execution and prediction operations are repeated.

[0112] Typical simulation environment and dynamic navigation verification

[0113] In the experiments, the obstacle avoidance performance of the three comparison algorithms was tested using three simulation environments with one to five dynamic obstacles, and the success rate of each algorithm was calculated. In this work, experimental success means successfully finding a feasible initial path and reaching the target point without collision. Experiment 3 involved 25 trials for each dynamic obstacle scenario, for a total of 125 trials. The purpose of the experiments was to demonstrate the reliability and robustness of the algorithms in dynamic environments, especially in highly dynamic scenarios.

[0114] like Figure 6 The following diagrams show three simulation environments: Sample (simple environment), Narrow (narrow passage), and Maze (maze structure), used to test the performance of the algorithm under different spatial complexities. The spatial dimensions of the Sample and Narrow scenes are both 100m×100m, and the Maze scene is 200m×200m.

[0115] Figure 7 Schematic diagrams of dynamic obstacle avoidance in three simulated environments demonstrate the performance of the proposed path planning method in a scenario with multiple dynamic obstacles. Obstacles move at varying speeds and directions within the scene, simulating the uncertain interference of pedestrians or equipment in real-world operations. The robot successfully completes the navigation task through the use of perception, prediction, and local repair mechanisms.

[0116] Figure 8 The following graph compares average execution cost and arrival time under three simulation environments, demonstrating the performance advantages of the P-RT-BFMT algorithm over RT-RRT* and RT-FMT under the same conditions. In experiments with varying numbers of dynamic obstacles ranging from 1 to 5, P-RT-BFMT demonstrated greater stability and efficiency in both execution cost and completion time. Figure 9 TABLE 1 shows the average success rates in the three simulation environments.

[0117] Time complexity and parameter adjustment

[0118] The P-RT-BFMT algorithm proposed in this paper fully considers the computing resource limitations and response delay issues during online execution in its design, and improves computing efficiency and real-time performance while ensuring path quality and dynamic adaptability.

[0119] (1) Time complexity analysis

[0120] During the initialization phase, the system uses the free space X free Perform collision-free sampling and construct a fixed sample set S. The generation process is a one-time calculation with a time complexity of

[0121] Bidirectional path expansion adopts heuristic bidirectional FMT strategy. Each node expansion involves nearest neighbor search and cost update. With the support of KD-Tree and other structures, the complexity of neighbor search can be controlled within The average time complexity of the overall path construction phase is

[0122] During the path execution process, the local path reconstruction operation is triggered only when a potential collision risk is detected, and the local planning process is limited to a small range within the heuristic domain;

[0123] The root node update mechanism records the path sequence index and tree structure reference relationship, eliminating the need to rebuild the path during node advancement.

[0124] In summary, the P-RT-BFMT algorithm designed in this application has good asymptotic computational efficiency and is particularly suitable for the deployment of navigation systems with real-time response requirements in dynamic environments.

[0125] (2) Key parameter adjustment mechanism

[0126] To improve the algorithm's versatility and adaptability to different platforms and tasks, the system's core parameters support flexible configuration. Commonly used parameters are recommended as follows:

[0127] Neighborhood radius:

[0128] Recommended s ∈[1.2,1.8]

[0129] Dynamic obstacle detection interval:

[0130] Δt=0.2~0.5s

[0131] Dynamic obstacle prediction time window length:

[0132] T = 1.5 to 2.5 seconds

[0133] The endpoint neighborhood radius and the heuristic domain radius can be set to:

[0134] r g =r h =0.5·r n

[0135] These parameters can be adjusted based on the computing power of the robot's platform, its sensing range, and the complexity of the task, achieving a good balance between response speed and path stability. The system provides a standard parameter interface and dynamic reconfiguration capabilities, supporting online parameter updates during mission execution.

[0136] Probabilistic Completeness Proof

[0137] In order to further verify the theoretical validity of the algorithm of the present invention, the following proves that it has probabilistic completeness, that is, when the number of samples tends to infinity, the algorithm can find the value from x in a probabilistic convergence way. s to x g a feasible path.

[0138] (1) Neighborhood radius

[0139] According to the PRM* theory, the following neighborhood radius is selected:

[0140]

[0141] d: workspace dimension;

[0142] γ s : regulatory factor;

[0143] N: total number of samples;

[0144] ·μ(X free ): the volume of free space;

[0145] ·ζ d is the volume of the unit sphere.

[0146] The radius is selected to ensure that as the number of samples N increases, the neighborhood radius r n The ability to reduce but not reduce spatial coverage ensures that enough sample points can be connected to construct a valid path.

[0147] (2) Sampling coverage and path approximation

[0148] If there is a true path x, the sampling process will generate a (∈, r n )-trace, that is, there is a sequence of path points satisfy:

[0149] The distance between adjacent path points satisfies ‖y m -y m+1 ‖≤r n ;

[0150] The approximate path cost satisfies c(y)≤(1+∈)c(x);

[0151] The distance between any real path point x(t) and the nearest point in the sampling point set does not exceed r n .

[0152] Define event A n "There is a valid path consisting of sample points connecting x s with x g ”.

[0153] According to random graph theory, failure events The probability of satisfying:

[0154]

[0155] As N→∞, the failure probability approaches 0 at a polynomial rate.

[0156] In summary, the P-RT-BFMT algorithm satisfies the probabilistic completeness. Its neighborhood connection radius r n This ensures sufficient coverage of free space, while the increased number of samples ensures path accessibility. Therefore, in theory, this is consistent with methods such as PRM*, FMT*, and BFMT*, and is a probabilistically complete sampling path planning algorithm.

[0157] The embodiments described above are merely descriptions of preferred implementations of the present invention and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should fall within the scope of protection determined by the claims of the present invention.

Claims

1. A robot real-time path planning system based on a predictive trigger mechanism, characterized by: It includes two groups of parallel modules, one group is a map processing module and a path construction module connected in sequence, and the other group is a dynamic obstacle perception module, a collision risk prediction module and a path repair module connected in sequence. The path construction module and the path repair module are simultaneously connected to the path update module.

2. A robot real-time path planning system based on a prediction trigger mechanism according to claim 1, characterized in that: The specific composition and functions of each module are as follows: Map processing module: responsible for loading the map data of the robot's working environment, preprocessing it, namely grayscale binarization and rasterization operations, and finally extracting the static obstacle space set X Fobs With Free Space Collection X free ; Path construction module: Take the starting point and the target point as the root and construct the forward tree T f and the backward tree T r , bidirectional FMT* algorithm is used for path expansion and connection to accelerate the path search process; Dynamic obstacle perception module: continuously obtains the position and speed information of dynamic obstacles around the robot and records them in real time to support subsequent trajectory prediction processing; Collision risk prediction module: This module estimates the future movement trajectory of dynamic obstacles based on their historical positions and velocities, and determines whether they may collide with the robot's current path. Path Repair Module: When the system predicts an impending collision, it triggers local path reconstruction and re-searches and repairs only the blocked path locations, improving the algorithm's local responsiveness and computational efficiency. Path update module: controls the robot to move along the currently planned path, and updates the root node, subpath, and execution cost in real time during the process to ensure the consistency and optimality of navigation execution.

3. A real-time robot path planning method based on a predictive trigger mechanism, implemented based on the planning system of claim 1 or 2, characterized in that: The specific steps are as follows: Step 1: Map loading and processing: The system loads the raster map or other environmental mapping data, obtains the static obstacle space information and free space area through binarization processing, and forms a basic navigation environment model; Step 2: Sample point generation and path tree initialization: Use the SampleFree algorithm to randomly and uniformly sample N collision-free nodes in free space to construct a spatial sampling graph for path planning, reducing the amount of repeated sampling and collision detection calculations during runtime; Step 3: Bidirectional path tree construction: The starting point and the end point are used as the root nodes of the forward tree and the backward tree, respectively. The bidirectional fast marching tree (BFMT) expansion mechanism is used to expand the path in each direction. When the distance between the nodes of the two trees is less than the set connection threshold, the path connection and merging operations are performed to generate the initial feasible path. Step 4: Path execution and root node update: The robot moves along the planned path step by step. Whenever the robot reaches a key node on the path, it automatically sets the node as the new path tree root node and updates the path structure to ensure path coherence and cost optimality. Step 5: Dynamic obstacle perception and trajectory prediction: When a moving obstacle enters the robot's perception range, the system initiates motion trajectory tracking and linear velocity estimation to predict and analyze the obstacle's future position. Step 6: Collision risk detection and local path repair: The system compares the predicted obstacle trajectory with the current robot's expected path in time and space. If there is a potential collision risk, the local path reconstruction mechanism is triggered to regenerate a collision-free feasible path in the affected area and replace the original path segment. Step 7: End point detection and task completion: When the robot enters the neighborhood of the end point, the system determines that the navigation task is completed and ends the path tracking execution.

4. A robot real-time path planning method based on a prediction trigger mechanism according to claim 3, characterized in that: Step 1 The specific steps are as follows: Step 1.1: Load map information from SLAM or predefined CAD drawings; Step 1.2: Perform a binarization operation on the map to identify the static obstacle area X Fobs Free movement area X free ; Step 1.3: Divide the free space into discrete computing units to provide basic data structure support for path search and collision detection.

5. The method for real-time robot path planning based on a prediction trigger mechanism according to claim 3, characterized in that: Step 2: Step 2.1: Call the SampleFree(N) algorithm, at X free Generate N collision-free random sample points; Step 2.2: Set the starting point x s is the forward tree T f The root node, target point x g is the backward tree T r The root node of Step 2.3: Initialize the open node set: V f-open ={x s },V r-open ={x g }。 6. The method for real-time robot path planning based on a prediction trigger mechanism according to claim 3, characterized in that: Step 3: Step 3.1: The system expands the forward tree T alternately or in parallel in each iteration f and the backward tree T r ; Step 3.2: Use the following heuristic cost function to guide the node expansion direction: Heuristic(x)=c(x)+||x-x g ‖ When the Euclidean distance between any two nodes in the two trees is less than the threshold δ, the path connection operation TreeMerge is performed.

7. The method for real-time robot path planning based on a prediction trigger mechanism according to claim 3, characterized in that: Step 4: The specific steps are as follows: Step 4.1: The robot moves along the currently connected path segment by segment. When the robot reaches the path node x k , set it as the new root node; Step 4.2: Trigger the path cost recalculation mechanism and update the path costs of its descendant child nodes: c(x q )=c(x p )+||x q -x p ||,x p =parent(x q ) Among them, c(x p ) refers to the current node cost, ||x q -x p || refers to the Euclidean distance between two nodes, x p is x q The parent node of .

8. The method for real-time robot path planning based on a prediction trigger mechanism according to claim 3, characterized in that: Step 5: The specific steps are as follows: Every Δt seconds, obtain the position information L of the dynamic obstacle at time t1 and time t2 t1 , L t2 , estimate its speed: Predict its future position sequence within the T=2s time window: At the same time, calculate the predicted trajectory Y of the robot within the path segment robot (t), if the following conditions are met: ||Y robot (t)-Y Dobs (t)||<R s This indicates that there is a potential conflict in the path, and the system triggers the local path reconstruction mechanism.

9. The method for real-time robot path planning based on a prediction trigger mechanism according to claim 3, characterized in that: Step 6: Add all blocked path nodes to the queue to be repaired M O ; Re-execute the Near neighbor node selection and Collisionless collision check operations in the local area, find an alternative parent node, build a new sub-path segment and replace the original path to complete the local reconstruction.

10. The method for real-time robot path planning based on a prediction trigger mechanism according to claim 3, characterized in that: Step 7: Determine whether the robot's current position satisfies: ||x robot -x g ||<=r g If the path has entered the neighborhood of the end point, the task is considered completed; otherwise, the path execution and prediction operations are repeated.

Images