A robot intelligent path planning method in unknown dynamic environment

Abstract

The present invention discloses a robot path planning method in an unknown dynamic environment. The first-level planner generates a preliminary planning path. The second-level planner establishes a dynamic obstacle constraint set for the dynamic obstacle collision that appears in the preliminary planning path, and starts one step before the collision. With the constraint set as the restriction condition, the first-level planner is incrementally called to continue the path planning. If a new dynamic obstacle collision occurs, its time step and coordinates are added to the set. If the time step has passed and the obstacle coordinates are not multi-order inner neighbors of the robot's current coordinates, the dynamic obstacle constraint is removed from the set, and the cycle is executed until the elements in the dynamic obstacle set are cleared. The present invention realizes adaptive intelligent real-time pathfinding of the robot in an unknown dynamic environment. Through the two-level planner, the robot's adaptability to various unknown dynamic environments is enhanced, and the real-time performance and success rate of the dynamic path planning method are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of robot dynamic path planning, and more specifically, to a robot intelligent path planning method in an unknown dynamic environment. Background Art

[0002] Path planning refers to planning a collision-free, feasible path connecting the starting point to the end point for the robot in an environment with unknown obstacles. At the same time, it should also meet one or more optimization indicators, such as path length, continuity, planning time, etc.

[0003] The existing dynamic environment path planning methods include hybrid algorithms and local path planning methods. The hybrid planning algorithm first divides the obstacles in the environment into static obstacles and dynamic obstacles, and then divides the planning task into a global planning stage and a re-planning stage. The global planning stage does not consider the dynamic obstacles in the environment. The existing A* algorithm, neural network and a large number of intelligent optimization algorithms can be used for global path planning; the re-planning stage is mainly when the robot encounters dynamic obstacles during the movement, the optimization algorithm is used to re-plan a feasible path to avoid dynamic obstacles, and the two cooperate with each other to guide the robot to move to the target point. The local path planning methods mainly include artificial potential field method, behavior decomposition method, dynamic window planning method, etc., but due to the lack of global information guidance, these local planning methods cannot guarantee that the planned path can reach the target point, nor can they guarantee the optimality of the planned path. In addition, local planning focuses on whether it can successfully avoid obstacles, and it is easy to fall into local optimality in complex environments, and the integrity of the planned path cannot be guaranteed. The above methods have shortcomings such as poor robustness, low efficiency and poor stability when facing complex environments. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a robot intelligent path planning method in an unknown dynamic environment, which has strong real-time performance, good stability, strong adaptability to the environment, and greatly improves the execution efficiency of the path planning method.

[0005] The technical solution of this application is as follows:

[0006] A robot intelligent path planning method in an unknown dynamic environment, characterized by comprising the following steps:

[0007] Step S1. Map environment modeling: divide grids into free grids and obstacle grids according to the distribution of obstacles in the environment. Before accessing the grids, mark the grids first;

[0008] Step S2. In the grid environment, the robot performs path planning, calculates the shortest path, and determines whether it has reached the end point. If not, execute step S3; if yes, execute step S5;

[0009] Step S3. Find the neighbor with the lowest cost and determine whether it is in the dynamic obstacle set. If so, wait in place. If not, move to the neighbor.

[0010] Step S4. Detect whether the obstacle has changed. If so, update the estimated cost correction value k m And set a new starting point, update the one-step-ahead objective function value rhs(s) of the multi-order neighbor, add it to the priority queue, recalculate the shortest distance, and return to step S2;

[0011] Step S5. Determine whether there is an obstacle collision, if yes, execute step S6, if no, end;

[0012] Step S6. Determine whether the dynamic obstacle collision coordinates are the same. If so, set the coordinates as temporarily unreachable and update the map. If not, execute step S7;

[0013] Step S7. Determine whether the previous dynamic obstacle is still there. If so, return to the time step before the collision and return to step S2; if not, set the coordinates to completely reach the updated map.

[0014] Furthermore, in step S2, before the robot takes the next step of walking, it needs to perform a priority evaluation on the surrounding neighbor grids, use the evaluation function to evaluate its surrounding grids, and select the grid with the smallest evaluation value as the next forward grid, the expression is f(s)=g(s)+h(s).

[0015] g(s) represents the actual cost from the end grid to the current position s, and h(s) represents the estimated cost from the current position s to the starting grid. h(s) is expressed using the Manhattan distance between the two grids, expressed as:

[0016] d(s,s')=|x s -x s' |+|y s -y s' |

[0017] Where (x s ,y s ) and (x s' ,y s' ) are the row and column coordinates of grid s and grid s' respectively.

[0018] Furthermore, the step S2 specifically includes setting the initial rhs(s) value and g(s) value of all grids in the map environment to positive infinity; setting the end point rhs(s) = 0; setting k m =0, dynamic obstacle constraint set D = 0, planning path set P = 0; set priority queue U = 0, insert the end point into priority queue U, and calculate the next step grid priority function value Key value of the end point;

[0019] Scan multiple order neighbors, if an obstacle change is detected, modify k m value, according to the heuristic value of the previous starting point and the current point, and set the current point as the previous starting point; continue to execute steps S3 and S5.

[0020] Furthermore, in step S3, the minimum rhs(s) among all neighboring points of the current position s is taken as the rhs value of s; if the current position s is in the priority queue, s is removed from the priority queue U; if g(s)!=rhs(s), s and its next step grid priority function value Key value are inserted into the priority queue.

[0021] Furthermore, when the priority queue U is not empty, and the minimum value of the priority queue is less than the next grid priority function value Key value of the grid node where the current robot is located, or the next grid priority function value Key value of the grid node where the current robot is located is inconsistent, execute the following loop: take the key value k_old of the highest priority point in queue U; pop the grid point u with the smallest next grid priority function value Key value in queue U; if the next grid priority function value Key value of u is greater than k_old, u is re-joined to the queue; if g(u)>rhs(u), set g=rhs and execute step S3; otherwise set g(u) to positive infinity, re-add u to the queue and execute step S3.

[0022] Furthermore, in step S4, when an obstacle appears and changes the map environment, g(s) needs to be updated in real time by the rhs(s) function. rhs(s) records the g(s) function value of the parent node of grid s plus the minimum value of the cost between these two points, which is equivalent to the minimum cost for a point to reach itself from the parent grid. The rhs(s) function expression is:

[0023] Among them, g(s') represents the actual cost from the end grid to grid s', s start is the grid node where the current robot is located, and Neighbors is the neighboring node of the grid node where the current robot is located.

[0024] Furthermore, in step S4, when a grid may need to be detected, it needs to be added to the queue U to be detected, in which the nodes to be detected and updated are stored according to the value of the grid priority function Key in the next step; for the same grid, different Keys will be generated in different situations, and Key will become the basis for the order of node detection when the first-level planner finds the path, including the first priority calculation function k1(s) and the second priority calculation function k2(s), first determine the value of the first priority calculation function k1(s), if the first priority calculation function k1(s) is equal, then determine the second priority calculation function k2(s), the specific formula is as follows:

[0025] k1(s)=min(g(s),rhs(s))+h(s start ,s)+k m ;

[0026] k2(s)=min(g(s),rhs(s))

[0027] in

[0028]

[0029] k m is h(s start ,s), h(s start ,s) is the estimated cost calculated in advance. Whenever the robot finds an obstacle change that is inconsistent with the previous environment, it will calculate h(s,s') based on the actual situation of the neighboring grid in the current map environment and perform a m The k1(s) calculated after the correction is accumulated makes the grid priority order in the queue U more reasonable. start is the current position of the robot, which is constantly changing. Therefore, the optimal key value in the previous environment is k_old. When the environment is updated, k can be used directly for faster calculation efficiency. m to correct k_old.

[0030] Furthermore, for the grids in queue U, their status is checked in the order of the next grid priority function value Key from small to large. After the detection, the state of the grid's multi-order neighbor is updated to determine whether the queue U needs to be performed again. The specific status is as follows:

[0031] (1) When a grid g(s) = rhs(s), the node in this state will not be included in the queue U;

[0032] (2) When a grid g(s)>rhs(s), set g(s)=rhs(s) and remove it from the queue U;

[0033] (3) When a grid g(s)<rhs(s), set g(s)=∞ and update its multi-order inner neighbors, and add it to the queue U again for further judgment.

[0034] Furthermore, in step S5, specifically, when the current node is not the end point, find the neighbor node set of the starting point; find the node corresponding to the minimum rhs value among the neighbors; if the point is not in set D, use the point as the new starting point; add the point to set P; and execute step S2.

[0035] Furthermore, the steps S6 and S7 are as follows: obtain the time and coordinates of the conflict in step S5, and add them to set D; if D=0, exit and return to set P; if there is a conflict, compare the first conflict grid coordinates in D with all the conflict coordinates in the previous conflict set D, and if they are equal, set the coordinates as temporarily unreachable; set the coordinates in Dd to be completely reachable; start from the time step before the first conflict, continue to execute step S5, and finally set P saves the planned path set.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] This method uses a two-level planner to plan the path. The first-level planner is an improved planning algorithm based on D*lite, which incrementally finds the path in the current environment. Once the robot predicts a collision, it does not need to re-plan from the beginning. On this basis, the waiting and detour decisions of the robot in the local environment are added; the second-level planner draws on the idea of ​​conflict search and increases the robot's judgment and obstacle avoidance of a large number of moving obstacles in unknown states. The robot realizes adaptive intelligent real-time pathfinding in an unknown dynamic environment. Through the two-level planner, the robot's adaptability to various unknown dynamic environments is enhanced, and the real-time and success rate of the dynamic path planning method are greatly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a flow chart of the present invention.

[0039] Figure 2 It is a schematic diagram of the grille identification of the present invention.

[0040] Figure 3 Schematic diagram of the third-order neighbors of the robot of the present invention. DETAILED DESCRIPTION

[0041] The technical solution of the present invention is further specifically described below through embodiments and in conjunction with the accompanying drawings.

[0042] like Figure 1 As shown, this embodiment provides a robot intelligent path planning method in an unknown dynamic environment, characterized by comprising the following steps:

[0043] Step S1. Map environment modeling: divide grids into free grids and obstacle grids according to the distribution of obstacles in the environment. Before accessing the grids, mark the grids first;

[0044] Step S2. In the grid environment, the robot performs path planning, calculates the shortest path, and determines whether it has reached the end point. If not, execute step S3; if yes, execute step S5;

[0045] Step S3. Find the neighbor with the lowest cost and determine whether it is in the dynamic obstacle set. If so, wait in place. If not, move to the neighbor.

[0046] Step S4. Detect whether the obstacle has changed. If so, update the estimated cost correction value k m And set a new starting point, update the one-step-ahead objective function value rhs(s) of the multi-order neighbor, add it to the priority queue, recalculate the shortest distance, and return to step S2;

[0047] Step S5. Determine whether there is an obstacle collision, if yes, execute step S6, if no, end;

[0048] Step S6. Determine whether the dynamic obstacle collision coordinates are the same. If so, set the coordinates as temporarily unreachable and update the map. If not, execute step S7;

[0049] Step S7. Determine whether the previous dynamic obstacle is still there. If so, return to the time step before the collision and return to step S2; if not, set the coordinates to completely reach the updated map.

[0050] In a further preferred embodiment, step S1 specifically includes: based on graph theory, a fixed-size grid is used to divide the robot workspace into multiple simple work areas, and then these grids are combined into a connected graph. According to the distribution of obstacles in the environment, the grid can be divided into two types: free grid and obstacle grid. A free grid means that the robot can walk freely in the grid, and an obstacle grid means that the grid is occupied by an obstacle and cannot be reached. For obstacles with irregular shapes, they are first segmented according to the grid size. The obstacle can occupy one or more grids. If the segmented obstacle does not completely occupy a grid, it is also regarded as a complete obstacle grid for the convenience of path planning.

[0051] Before accessing the grid, you need to mark the grid first. The grid map includes feasible grids (fully accessible) and barrier grids (no access). Figure 2 As shown, a two-dimensional rectangular coordinate system is established with the lower left corner as the origin. The black grid is the obstacle grid, the white grid is the free grid, S represents the starting point, and G represents the end point. The grids can be numbered according to their row and column coordinates.

[0052] like Figure 3As shown in the figure, assuming that robot o is located at the center of the grid and the robot is limited to 8 degrees of freedom movement, then the next time node may have 8 movement directions, including upper left, upper, upper right, left, right, lower left, lower and lower right. With robot o as the center of the grid, from inside to outside are o's first-order neighbors, second-order neighbors and third-order neighbors. The robot will incur costs when walking between grids. Since the grids are of the same size, it is set that the robot needs to spend 1 unit cost when walking between two adjacent grids. For walking in the upper left, upper right, lower left and lower right diagonal grids, the robot will spend 1.4 units of cost. If the neighbor is an obstacle grid, the cost is infinite.

[0053] In a further preferred embodiment, in step S2, before the robot takes the next step of walking, it needs to perform a priority evaluation on the surrounding neighbor grids, use an evaluation function to evaluate its surrounding grids, and select the grid with the smallest evaluation value as the next forward grid, expressed as f(s)=g(s)+h(s).

[0054] g(s) represents the actual cost from the end grid to the current position s, and h(s) represents the estimated cost from the current position s to the starting grid. h(s) is expressed using the Manhattan distance between the two grids, expressed as:

[0055] d(s,s')=|x s -x s' |+|y s -y s' |

[0056] Where (x s ,y s ) and (x s' ,y s' ) are the row and column coordinates of grid s and grid s' respectively.

[0057] When an obstacle appears and changes the map environment, g(s) needs the rhs(s) function to be updated in real time. rhs(s) records the g(s) function value of the parent node of grid s plus the minimum value of the cost between these two points.

[0058] When a point reaches the minimum cost from the parent grid to itself, the rhs(s) function expression is:

[0059]

[0060] Among them, g(s') represents the actual cost from the end grid to grid s', s start is the grid node where the current robot is located, and Neighbors is the neighboring node of the grid node where the current robot is located. start Neighbors

[0061] When a grid may need to be detected, it needs to be added to the queue U for detection, which stores the nodes to be detected and updated according to the value of the grid priority function Key in the next step; for the same grid, different Keys will be generated in different situations. Key will become the basis for the order of node detection when the first-level planner finds the path, including the first priority calculation function k1(s) and the second priority calculation function k2(s). First, the value of the first priority calculation function k1(s) is determined. If the first priority calculation function k1(s) is equal, then the second priority calculation function k2(s) is determined. The specific formula is as follows:

[0062] k1(s)=min(g(s),rhs(s))+h(s start ,s)+k m ;

[0063] k2(s)=min(g(s),rhs(s))

[0064] in

[0065]

[0066] Among them, goal is the end point, otherwise is the other point except the end point, k m is h(s start ,s), h(s start ,s) is the estimated cost calculated in advance. Whenever the robot finds an obstacle change that is inconsistent with the previous environment, it will calculate h(s,s') based on the actual situation of the neighboring grid in the current map environment and perform a m The k1(s) calculated after accumulation and correction makes the grid priority order in the queue U more reasonable.

[0067] For the grids in queue U, the algorithm will check their status in the order of the next grid priority function value Key from small to large. After the detection, the state of the grid's third-order neighbors will be updated to determine whether it is necessary to queue U again. The specific status is as follows:

[0068] 1. When a grid g(s) = rhs(s), it means that the grid is in a stable state with no error between the estimated and actual values. If the external environment does not change, this state will be maintained. Nodes in this state will not be included in the queue U.

[0069] 2. When a grid g(s)>rhs(s), it means that the edge cost function value becomes smaller, which means that the obstacles on the grid are cleared or the algorithm can search for a better path. When a grid is detected in this state, set g(s)=rhs(s) and remove it from the queue U.

[0070] 3. When a grid g(s) < rhs(s), it means that there may be new obstacles near the grid, and the cost becomes larger. When a grid is detected in this state, set g(s) = ∞ and update its third-order inner neighbors, and add it to U again for further judgment.

[0071] The first-level planner will generate a preliminary planning path. There may be dynamic obstacle collisions in the path. For these dynamic obstacles, the second-level planner will establish a dynamic obstacle constraint set, add the time steps and coordinates of these dynamic obstacle collisions to the set, and start one step before the collision. With the constraint set as the restriction condition, the first-level planner will be incrementally called to continue path planning. If a new dynamic obstacle collision occurs, its time step and coordinates will be added to the set. If the time step has passed and the obstacle coordinates are not the third-order inner neighbors of the robot's current coordinates, the dynamic obstacle constraint will be removed from the set. This cycle will be executed until the elements in the dynamic obstacle set are cleared. At this point, the second-level planner will generate a collision-free planning path.

[0072] The specific process is as follows:

[0073] Set the initial rhs(s) and g(s) values ​​of all grids in the map environment to positive infinity; set the end point rhs(s) = 0; set k m =0, dynamic obstacle constraint set D = 0, planning path set P = 0; set priority queue U = 0, insert the end point into priority queue U, and calculate the next step grid priority function value Key value of the end point;

[0074] Scan the neighbors within the 3rd order. If an obstacle change is detected, modify k m value, according to the heuristic value of the previous starting point and the current point, and set the current point as the previous starting point; continue to execute steps S3 and S5.

[0075] In step S3, the minimum rhs(s) among all neighboring points of the current position s is taken as the rhs value of s; if the current position s is in the priority queue, s is removed from the priority queue U; if g(s)!=rhs(s), s and its next step grid priority function value Key value are inserted into the priority queue.

[0076] When the priority queue U is not empty, and the minimum value of the priority queue is less than the next grid priority function value Key value of the grid node where the current robot is located, or the next grid priority function value Key value of the grid node where the current robot is located is inconsistent, execute the following loop: take the key value k_old of the highest priority point in queue U; pop the grid point u with the smallest next grid priority function value Key value in queue U; if the next grid priority function value Key value of u is greater than k_old, u rejoins the queue; if g(u)>rhs(u), set g=rhs and execute step S3; otherwise set g(u) to positive infinity.

[0077] In step S5, specifically, when the current node is not the end point, find the neighbor node set of the starting point; find the node corresponding to the minimum rhs value among the neighbors; if the point is not in set D, use the point as the new starting point; add the point to set P; and execute step S2.

[0078] Specifically, steps S6 and S7 obtain the time and coordinates of the conflict in step S5 and add them to set D; if D=0, exit and return to set P; if there is a conflict, compare the first conflict grid coordinates in D with all the conflict coordinates in the previous conflict set D. If they are equal, set the coordinates as temporarily unreachable; set the coordinates in Dd as fully reachable; start from the time step before the first conflict, continue to execute step S5, and finally set P saves the planned path set.

[0079] The specific embodiments described herein are merely examples of the spirit of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in similar ways, but they will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

Claims

1. A robot intelligent path planning method in an unknown dynamic environment, characterized in that: It includes the following steps: Step S1. Map environment modeling: According to the distribution of obstacles in the environment, the grid is divided into free grids and obstacle grids. Before accessing the grid, the grid is first identified; Step S2. In the grid environment, the robot performs path planning, calculates the shortest path, and determines whether it has reached the end point. If not, go to step S3; if so, go to step S5; Step S3. Find the neighbor with the smallest cost, and determine whether it is in the dynamic obstacle set. If so, wait in place; if not, move to that neighbor; Step S4. Detect whether the obstacle has changed. If so, update the estimated cost correction value k m And set a new starting point, update the one-step-ahead objective function value rhs(s) of the multi-order neighbor, add it to the priority queue, recalculate the shortest distance, and return to step S2; When an obstacle appears during the process and changes the map environment, the g(s) needs to be updated in real time by the rhs(s) function. The rhs(s) records the minimum value of the g(s) function of the parent node of the grid s plus the cost between these two points, which is equivalent to the minimum cost for a point to reach itself from the parent grid. The expression of the rhs(s) function is: Among them, g(s') represents the actual cost from the end grid to grid s', s start is the grid node where the current robot is located, Neighbors is the neighboring node of the grid node where the current robot is located; d(s, s′) represents the Manhattan distance between the end grid and the grid s′; When a grid may need to be detected, it needs to be added to the detection queue U, which stores the nodes to be detected and updated in ascending order of the Key value; for the same grid, different Key values will be generated in different situations. The Key will be the basis for the order of node detection when the first-level planner searches for a path, including the first priority calculation function k1(s) and the second priority calculation function k2(s). First, judge the value of the first priority calculation function k1(s). If the first priority calculation function k1(s) is equal, then judge the second priority calculation function k2(s). The specific formulas are as follows: k1(s)=min(g(s),rhs(s))+h(s start ,s)+k m ; k2(s) = min(g(s), rhs(s)) where k m is h(s start ,s), h(s start ,s) is the estimated cost calculated in advance. Whenever the robot finds an obstacle change that is inconsistent with the previous environment, it will calculate h(s,s') based on the actual situation of the neighboring grid in the current map environment and perform a m Accumulate and modify the calculated k1(s) to make the grid priority order in the queue U more reasonable. goal represents the target grid. For the grids in the queue U, their statuses are detected in ascending order of the Key value. After the detection, the statuses of the multi-level inner neighbors of the grid will be updated to determine whether the queue U needs to be entered again. The specific statuses are as follows: (1). When g(s) = rhs(s) for a grid, the node in this state will not be included in the queue U; (2). When g(s) > rhs(s) for a grid, let g(s) = rhs(s) and remove it from the queue U; (3). When g(s) < rhs(s) for a grid, let g(s) = ∞ and update its multi-level inner neighbors, and at the same time add it to the queue U again for further judgment; Step S5. Determine whether there is an obstacle collision. If so, go to step S6; if not, end; Step S6. Determine whether the dynamic obstacle collision coordinates are the same. If so, set this coordinate as temporarily unreachable and update the map; if not, go to step S7; Step S7. Determine whether the previous dynamic obstacle still exists. If so, return to the time step before the collision and go back to step S2; if not, set this coordinate as fully reachable and update the map.

2. According to the method of claim 1, the robot intelligent path planning method in an unknown dynamic environment is characterized by: In step S2, before the robot walks in the next step, it needs to evaluate the priorities of the surrounding neighbor grids, evaluate its surrounding grids using the evaluation function, and select the grid with the smallest evaluation value as the next forward grid. The expression is f(s) = g(s) + h(s); g(s) represents the actual cost from the end grid to the current position s, and h(s) represents the estimated cost from the current position s to the starting grid. h(s) is expressed using the Manhattan distance between the two grids, expressed as: d(s,s')=|x s -x s' |+|y s -y s' | Where (x s ,y s ) and (x s' ,y s' ) are the row and column coordinates of grid s and grid s' respectively.

3. The method for intelligent path planning of a robot in an unknown dynamic environment according to claim 2, characterized in that: The step S2 specifically includes setting the initial rhs(s) value and g(s) value of all grids in the map environment to positive infinity; setting the end point rhs(s)=0; setting k m =0, dynamic obstacle constraint set D = 0, planning path set P = 0; set priority queue U = 0, insert the end point into priority queue U, and calculate the next step grid priority function value Key of the end point; Scan multiple order neighbors, if an obstacle change is detected, modify k m Value, based on the heuristic value of the previous starting point and the current point, and set the current point to the previous starting point; Continue to execute steps S3 and S5.

4. The method for intelligent path planning of a robot in an unknown dynamic environment according to claim 3, characterized in that: In step S3, the minimum rhs(s) among all neighboring points of the current position s is taken as the rhs value of s; if the current position s is in the priority queue, s is removed from the priority queue U; if g(s)!=rhs(s), s and its next step grid priority function value key are inserted into the priority queue.

5. The method for intelligent path planning of a robot in an unknown dynamic environment according to claim 4, characterized in that: When the priority queue U is not empty, and the minimum value of the priority queue is less than the next grid priority function value key of the grid node where the current robot is located, or the next grid priority function value key of the grid node where the current robot is located is inconsistent, execute the following loop: take the minimum value k_old of the next grid priority function value key of the highest priority point in queue U; pop the grid point u with the smallest next grid priority function value Key in queue U; if the next grid priority function value key of u is greater than k_old, u is re-added to the queue; if g(u)>rhs(u), set g=rhs and execute step S3; otherwise set g(u) to positive infinity, re-add u to the queue and execute step S3.

6. The method for intelligent path planning of a robot in an unknown dynamic environment according to claim 1, characterized in that: In step S5, specifically, when the current node is not the end point, find the neighbor node set of the starting point; find the node corresponding to the minimum rhs value among the neighbors; if the point is not in the set D, use the point as the new starting point; Add the point to the set P; execute step S2.

7. The method for intelligent path planning of a robot in an unknown dynamic environment according to claim 1, characterized in that: Specifically, step S6 and step S7 are as follows: obtain the time and coordinates of the conflict in step S5, and add them to set D; if D=0, exit and return to set P; if there is a conflict, compare the first conflict grid coordinates in D with all the conflict coordinates in the previous conflict set D, and if they are equal, set the coordinates as temporarily unreachable; set the coordinates in Dd as fully reachable; start from the time step before the first conflict, continue to execute step S5, and finally set P saves the planned path set.

Images