WO2023155923A1 - 路径规划方法及装置、可读存储介质 - Google Patents

路径规划方法及装置、可读存储介质 Download PDF

Info

Publication number
WO2023155923A1
WO2023155923A1 PCT/CN2023/077432 CN2023077432W WO2023155923A1 WO 2023155923 A1 WO2023155923 A1 WO 2023155923A1 CN 2023077432 W CN2023077432 W CN 2023077432W WO 2023155923 A1 WO2023155923 A1 WO 2023155923A1
Authority
WO
WIPO (PCT)
Prior art keywords
tree
node
tree node
target
search
Prior art date
Application number
PCT/CN2023/077432
Other languages
English (en)
French (fr)
Inventor
吴丹
辛睿山
Original Assignee
中兴通讯股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中兴通讯股份有限公司 filed Critical 中兴通讯股份有限公司
Publication of WO2023155923A1 publication Critical patent/WO2023155923A1/zh

Links

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J9/00—Programme-controlled manipulators
    • B25J9/16—Programme controls
    • B25J9/1656—Programme controls characterised by programming, planning systems for manipulators
    • B25J9/1664—Programme controls characterised by programming, planning systems for manipulators characterised by motion, path, trajectory planning
    • B25J9/1666—Avoiding collision or forbidden zones
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J9/00—Programme-controlled manipulators
    • B25J9/16—Programme controls
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B25—HAND TOOLS; PORTABLE POWER-DRIVEN TOOLS; MANIPULATORS
    • B25J—MANIPULATORS; CHAMBERS PROVIDED WITH MANIPULATION DEVICES
    • B25J9/00—Programme-controlled manipulators
    • B25J9/16—Programme controls
    • B25J9/1602—Programme controls characterised by the control system, structure, architecture
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
    • G01C21/20—Instruments for performing navigational calculations
    • G—PHYSICS
    • G05—CONTROLLING; REGULATING
    • G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
    • G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
    • G05D1/02—Control of position or course in two dimensions

Definitions

  • the embodiments of the present application relate to the technical field of route navigation, and in particular, to a route planning method and device, and a computer-readable storage medium.
  • the path planning of moving objects is an important part of industrial control. It mainly refers to finding a continuous collision-free path from the starting point to the target point in the pose space under the condition of meeting the environmental constraints and the requirements of the moving object itself.
  • the mainstream path planning algorithm is mainly a sampling-based planning algorithm represented by Rapid-exploration Random Tree (RRT). This algorithm has strong randomness but takes a long time to construct the path, which cannot meet the current path planning requirements. .
  • RRT Rapid-exploration Random Tree
  • Embodiments of the present application provide a path planning method and device, and a computer-readable storage medium.
  • the embodiment of the present application provides a path planning method, including: obtaining the motion space, starting point position, end point position and obstacle information of the moving object; A search tree, and constructing a second search tree from the end position, wherein the first search tree and the second search tree grow oppositely according to the obstacle information; when the first search tree grows The distance between the first tree node and the second tree node grown by the second search tree is not greater than a preset step size, connecting the first search tree and the second search tree to obtain the plan of the moving object path; wherein, the second tree node grows in the following manner: from the node set of the second exploration tree, select the tree node closest to the target sampling point as the third tree node, and select the first The tree node closest to the second search tree on the search tree is a target tree node, wherein the target sampling point is obtained by expanding sampling along the last grown second tree node; when the third tree node reaches The shortest distance of the obstacle is not less than the preset distance threshold, according to the direction of the direction of the
  • the embodiment of the present application also provides a path planning device, including: a memory, a processor, and a computer program stored on the memory and operable on the processor.
  • a path planning device including: a memory, a processor, and a computer program stored on the memory and operable on the processor.
  • the processor executes the computer program, the The path planning method described in the first aspect above.
  • the embodiment of the present application further provides a computer-readable storage medium storing computer-executable instructions, and the computer-executable instructions are used to execute the path planning method described in the first aspect above.
  • FIG. 1 is a flowchart of a path planning method provided by an embodiment of the present application
  • Fig. 2 is a flow chart of growing the first tree node in the path planning method provided by one embodiment of the present application
  • Fig. 3 is a flow chart of constructing a first search tree in the path planning method provided by an embodiment of the present application
  • FIG. 4 is a flow chart of building a first search tree in the path planning method provided by another embodiment of the present application.
  • FIG. 5 is a flow chart of growing a second tree node in the path planning method provided by an embodiment of the present application.
  • FIG. 6 is a flow chart of growing a second tree node in the path planning method provided by another embodiment of the present application.
  • FIG. 7 is a flow chart of building a second search tree in the path planning method provided by an embodiment of the present application.
  • FIG. 8 is a flow chart of building a second search tree in the path planning method provided by another embodiment of the present application.
  • FIG. 9 is a flow chart of obtaining a planned path of a moving object in the path planning method provided by an embodiment of the present application.
  • Fig. 10 is a schematic diagram of the growth and expansion of the first search tree and the second search tree provided by an embodiment of the present application;
  • FIG. 11 is a schematic diagram of a comparison between a two-dimensional path determined based on a path planning method provided by an embodiment of the present application and a two-dimensional path determined based on a path planning method in the related art;
  • Fig. 12 is a schematic diagram of a comparison between a three-dimensional path determined based on a path planning method provided by an embodiment of the present application and a three-dimensional path determined based on a path planning method in the related art;
  • Fig. 13 is a schematic diagram of a path planning device provided by an embodiment of the present application.
  • the present application provides a path planning method and device, and a computer-readable storage medium, which uses the way that the first search tree and the second search tree grow opposite each other to find the path planning space, and connects the first search tree and the second search tree.
  • Obtaining the planned path of the moving object can effectively improve the obstacle avoidance effect of the path construction, and consider the distance influence of the relative obstacle, and use the combination of random learning and target learning to construct the tree nodes of the exploration tree, which can balance the randomness of tree node growth. It can effectively reduce the possibility of collision between the generated tree nodes and obstacles, so that the exploration tree can converge to the target faster, thereby reducing the path planning time, and obtaining a more optimized non-collision continuous path, which can meet the path planning requirements.
  • Figure 1 is a flowchart of a path planning method provided by an embodiment of the present application, the path planning method Including but not limited to steps S100 to S300.
  • Step S100 Obtain the movement space, starting position, end position and obstacle information of the moving object.
  • the start position and end position of the moving object are respectively the start point and end point of the planned path to be constructed, which can be expressed in two-dimensional, three-dimensional or even more dimensions;
  • the motion space of the moving object represents the moving object
  • the activity space is used to constrain the overall range of the planned path to be constructed so that there will be no movement space beyond the moving object, which can be expressed in two, three or even more dimensions; obstacles correspond to moving objects , is the obstacle for the moving object to move.
  • One of the requirements of the designed planning path is to avoid obstacles. Therefore, by obtaining the obstacle information, it can provide the basis for the collision detection of the growing tree nodes, so as to design a plan that meets the requirements. path.
  • moving objects there are many types of moving objects, and the smart objects commonly used in the industrial field that need to be moved can be the moving objects in this embodiment, such as but not limited to robotic arms, robots, and machine-controlled objects. Cars, etc., mobile objects can be selected and set in relevant application scenarios, which is not limited in this embodiment.
  • Step S200 In the motion space, construct the first exploration tree from the start position, and construct the second exploration tree from the end position, wherein the first exploration tree and the second exploration tree grow opposite to each other according to the obstacle information;
  • the second tree node grown by the second exploration tree may be grown in the following manner, but not limited to:
  • a combination of random learning and target learning is used to construct the tree nodes of the exploration tree, which can balance the randomness and targetness of tree node growth, and effectively reduce the relationship between the generated tree nodes and The possible collision of obstacles enables the search tree to converge to the target faster, thereby reducing the path planning time and obtaining a more optimized non-collision continuous path, which can meet the path planning requirements.
  • the first exploration tree gradually expands and grows the first tree node along the starting point position toward the second exploration tree
  • the second exploration tree gradually expands and grows the second tree node along the end position toward the first exploration tree
  • the set of multiple first tree nodes is the node set of the first exploration tree
  • the set of multiple second tree nodes is the node set of the second exploration tree.
  • there can be multiple target sampling points and each target sampling point is obtained by expanding sampling along the second tree node that was grown last time.
  • the two exploration trees can continue to grow opposite to each other to achieve convergence. It can be seen as a whole growth tree that continues to grow until the relevant conditions are met, then it can be confirmed that the growth tree has grown, and then the planned path of the moving object can be obtained by means of the growth tree.
  • the growth can be started from the first search tree or from the second search tree, which is applicable to all embodiments of this application; after the second tree node is generated, if necessary, continue to To grow the exploration tree, the currently obtained second tree node is used as the basis for growing the next second tree node, that is, it can be carried out in a loop.
  • the current target sampling point is obtained by expanding sampling along the end position, then similarly, The next target sampling point can be obtained based on the expansion sampling of the currently obtained second tree node, and can also be grown in a similar way to obtain the next second tree node.
  • the first tree node can also be grown in a cyclic manner , will not be described in detail; through the cyclical alternate growth shown in the above process, the first search tree and the second search tree that meet the requirements can finally be obtained.
  • the target sampling point x rand is equal to the current target point, that is, the target tree node is selected as the target sampling point x rand , otherwise, the target The sampling point x rand is randomly selected within the scope of the motion space expanded along the second tree node grown last time, and the generation formula of the target sampling point x rand is as follows:
  • Uniform( ) is a uniform sampling function
  • P is a random number in [0, 1]
  • gais represents the target bias probability value, which can be set to 0.05, but not limited to, x goal is the target solution space, and x free is the moving object exercise space.
  • the current target node of the second search tree is the tree node closest to the second search tree on the first search tree, so the growth tree node of the first search tree can be used to determine the node of the second search tree.
  • the target tree node will be described in detail in the following specification.
  • the shortest distance from the third tree node to the obstacle represents the relative positional relationship between the corresponding nearest neighbor node and the obstacle; when the shortest distance from the third tree node to the obstacle is not less than the preset distance threshold, It indicates that the nearest neighbor node is located far away from the obstacle area.
  • the target learning aspect of the new growth node can be increased, that is, the target learning part along the direction of the third tree node towards the target tree node can be increased, and at the same time, the new growth node can be combined with
  • the random learning part along the direction of the third tree node toward the target sampling point is reserved, so that the new growth node quickly converges to the target, which is conducive to improving the convergence rate, so that the moving object can successfully avoid obstacles and quickly reach the target.
  • the values of the preset distance threshold and the first preset probability threshold are not limited, and can be set according to actual application scenarios, for example, but not limited to, considering setting based on obstacle information, movement space of moving objects, etc. This is not limited in this embodiment.
  • the first tree node grown by the first exploration tree may be grown in the following manner, but not limited to:
  • Step S400 From the node set of the first exploration tree, determine that the tree node closest to the starting sampling point is the fourth tree node, wherein the starting sampling point is obtained by expanding sampling along the first tree node grown last time;
  • Step S500 Along the direction of the fourth tree node toward the starting sampling point, a tree node is grown according to a preset step size as the first tree node.
  • the starting sampling point is obtained by sampling, and then from the node set of the first exploration tree, the nearest neighbor node of the starting sampling point is determined to be the fourth tree node, so that the direction of the starting sampling point from the fourth tree node can be The direction is randomly grown to obtain the corresponding first tree node, that is, the first tree node is determined by random growth, so that the constructed planning path has certain probability completeness.
  • the starting sampling point can be generated in the obstacle-free area, but not limited to, according to the following steps: generating the starting sampling probability according to a uniform probability distribution; when the starting sampling probability is less than the second preset probability threshold, from the second search tree Select the start tree node from the node set as the start sampling point, wherein the start tree node is the tree node closest to the first search tree on the second search tree; when the start sampling probability is greater than or equal to the second preset probability threshold, along the The first tree node expansion of a growth randomly selects the starting sampling point.
  • the acquisition method of the start sampling point is similar to the acquisition method of the target sampling point and the principle is the same. Since the foregoing embodiments have described the acquisition method of the target sampling point in detail, in order to avoid redundancy, it will not be repeated here. .
  • the preset step size is the length between two tree nodes grown sequentially on the exploration tree, and its value is not limited, and can be selected by those skilled in the art in actual application scenarios. In some scenarios, when Determine the default step size of After that, all tree nodes in the first search tree and the second search tree can grow according to the preset step size.
  • step S200 includes but not limited to step S210.
  • Step S210 When it is determined that there is a collision between the first tree node and the obstacle, the first tree node is discarded, and the second tree node of the second search tree is grown.
  • the first tree node when it is determined that there is a collision between the first tree node and an obstacle, it means that the currently growing first tree node does not meet the path requirements, so the first tree node is discarded, and the first tree node is not used as At the same time, a part of the first search tree is turned to grow the nodes of the second tree from the second search tree on the other side, so as to realize the alternate growth of the first search tree and the second search tree.
  • step S200 also includes but is not limited to step S220.
  • Step S220 When it is determined that the first tree node does not collide with the obstacle, add the first tree node to the node set of the first search tree.
  • the currently grown first tree node meets the path requirements, so the currently grown first tree node can be retained and added to In the node set of the first search tree, use it as a part of the first search tree, and continue to grow the first search tree along the first tree node until the first tree node collides with an obstacle, Growth and expansion of the first search tree can thus be achieved.
  • the first tree node collides with the obstacle. For example, it may first be judged whether the generated first tree node is within the position range of the obstacle, and if not, it may be Further judge whether the connection between the nearest neighbor point and the generated first tree node collides with an obstacle; if so, discard the generated first tree node and switch to the second search tree for growth, otherwise the first The tree nodes are added to the node set of the first exploration tree and become a part of the first exploration tree, that is, they are used as a part of the planned path.
  • step S200 “grow a tree node as the second tree node according to the direction of the third tree node toward the target sampling point and the direction of the third tree node toward the target tree node" in step S200 includes but does not It is limited to steps S230 to S260.
  • Step S230 growing a random learning vector along the direction of the third tree node towards the target sampling point;
  • Step S240 growing a target learning vector along the direction of the third tree node towards the target tree node;
  • Step S250 adding the random learning vector and the target learning vector to obtain a comprehensive learning vector
  • Step S260 Along the direction of the comprehensive learning vector, a tree node is grown according to a preset step size as a second tree node.
  • growing a random learning vector along the direction of the third tree node towards the target sampling point is equivalent to growing along the direction of the third tree node towards the target sampling point
  • a random learning node, while growing a target learning vector along the direction of the third tree node towards the target tree node is equivalent to growing a target learning node along the direction of the third tree node towards the target tree node, that is, combining random learning and target learning
  • the idea of growing the second tree node based on the third tree node not only inherits the advantages of strong randomness of RRT, but also has the advantage of fast convergence speed of PSO, so that the second exploration tree can grow to the target node more quickly and reduce the planning path length.
  • the generated second tree node not only takes into account the influence of the random learning vector, but also considers the influence of the target learning vector, so from the actual construction path, it appears as: the third tree node towards the second tree node Direction, located between the direction of the third tree node towards the target sampling point and the direction of the third tree node towards the target tree node. From this part, it can be seen that the distance between the nearest neighbor nodes between the two exploration trees will be shortened , so the planning path length can be reduced.
  • the second tree node can also be grown in the following manner, but not limited to:
  • Step S270 When the shortest distance from the third tree node to the obstacle is less than the preset distance threshold, a tree node is grown along the direction of the third tree node toward the target sampling point according to the preset step length as the second tree node.
  • the shortest distance from the third tree node to the obstacle represents the relative positional relationship between the corresponding nearest neighbor node and the obstacle; when the shortest distance from the third tree node to the obstacle is less than the preset distance threshold, then It indicates that the nearest neighbor node is located in the area close to the obstacle.
  • the random learning aspect of the new growth node can be increased, that is, the random learning part along the direction of the third tree node towards the target sampling point can be increased, so that the new growth node can quickly move to Target convergence is beneficial to improve the convergence rate, so that the moving object can successfully avoid obstacles and reach the target quickly.
  • step S200 includes but not limited to step S280.
  • Step S280 When it is determined that there is a collision between the second tree node and the obstacle, the second tree node is discarded, and the first tree node of the first search tree is grown.
  • the second tree node when it is determined that there is a collision between the second tree node and the obstacle, it means that the currently growing second tree node does not meet the path requirements, so the second tree node is discarded and the second tree node is not used as At the same time, a part of the second search tree is turned to grow the nodes of the first tree from the first search tree on the other side, so as to realize the alternate growth of the first search tree and the second search tree.
  • step S200 also includes but is not limited to step S290.
  • Step S290 When it is determined that the second tree node does not collide with the obstacle, add the second tree node to the node set of the second search tree.
  • the currently growing second tree node meets the path requirements, so the currently growing second tree node can be retained and added To the node set of the second exploration tree, use it as a part of the second exploration tree, and continue to grow the second exploration tree along the second tree node until the second tree node collides with an obstacle , so the growth and expansion of the second exploration tree can be realized.
  • the node is added to the node set of the second exploration tree and becomes a part of the second exploration tree, that is, it is used as a part of the planned path.
  • Step S300 When the distance between the first tree node grown in the first search tree and the second tree node grown in the second search tree is not greater than the preset step size, connect the first search tree and the second search tree to obtain the moving object planning path.
  • the first search tree and the second search tree grow oppositely to find the path planning space, and the planned path of the moving object is obtained by connecting the first search tree and the second search tree, which can effectively improve the construction
  • the tree nodes of the exploration tree are constructed by combining random learning and target learning, which can balance the randomness and target of tree node growth, and effectively reduce the number of generated tree nodes.
  • the possibility of collision with obstacles enables the search tree to converge to the target faster, thereby reducing the path planning time and obtaining a more optimized non-collision continuous path, which can meet the path planning requirements.
  • step S300 includes but not limited to step S310 .
  • Step S310 Connect the first search tree and the second search tree respectively according to the growth order of the tree nodes, and connect the first tree nodes With the second tree node, the planned path of the moving object is obtained.
  • each first node on the first search tree may be cyclically grown in sequence according to the expansion order.
  • the tree nodes are connected, and the second tree nodes on the second exploration tree are connected according to the order of expansion, and finally the nearest nodes of the two exploration trees, that is, the first tree node with the closest distance and the second tree node are connected.
  • the nodes of the two trees are connected to obtain an overall continuous collision-free planning path.
  • Figure 10 is a schematic diagram of the growth and expansion of the first search tree and the second search tree provided by an embodiment of the present application, wherein Figure 10 shows a schematic diagram under two-dimensional conditions, other three-dimensional or even more The situation of multi-dimensional conditions can be deduced in the same way, and this embodiment is only explained by two-dimensional extended growth
  • the path planning method provided by this application can be executed according to the following process:
  • random sampling can be based on the two-way rapid expansion random tree (Rapid-exploration Random Tree Connect, RRT-Connect) to add the target bias part, and the mechanical arm
  • RRT-Connect Rapid-exploration Random Tree Connect
  • the starting position of T1 is used as the starting node ⁇ T1,start of T1
  • ⁇ T1,start is added to the node set E1 of T1
  • the target position x goal is used as the starting node ⁇ T2, start of T2, and ⁇ T2,start is added to In the node set E2 of T2.
  • Growth tree Tree is the tree currently growing, first set Tree to T1.
  • the process of growing for T1 includes:
  • Step 1 Generate a starting sampling point in a collision-free environment, such as ⁇ T1,rand1 in Figure 10, and perform random sampling based on the target bias, that is, pre-generate a uniform probability P1.
  • start Sampling point ⁇ T1, rand1 is equal to the current target point, otherwise, start sampling point ⁇ T1, rand1 is randomly selected within the scope of the motion space along the first tree node of last growth;
  • x 1 and x 2 represent any two nodes in the Tree
  • ⁇ (x 1 , x 2 ) represents the distance between x 1 and x 2 ;
  • Step 3 ⁇ nearest1 generates a new node ⁇ T1 , new1 with a certain step size in the direction of ⁇ T1, rand1 ;
  • Step 4 Carry out collision detection on the new node ⁇ T1, new1 , if it is determined that the new node ⁇ T1, new1 does not collide with an obstacle, then add ⁇ T1, new1 to the node set of Tree, otherwise discard the node.
  • Step 5 Generate target sampling points in a collision-free environment, such as ⁇ T2,rand1 in Figure 10, and perform random sampling based on the target bias, that is, pre-generate a uniform probability P2.
  • the target Sampling point ⁇ T2, rand1 is equal to the current target point, otherwise, the target sampling point ⁇ T2, rand1 is randomly selected within the scope of the motion space along the second tree node expansion of last growth;
  • Step 6 as shown in Figure 10, in the node set of Tree, according to the Euclidean distance, find the nearest neighbor of distance ⁇ T2, rand1 Node ⁇ nearest2 , since the nearest neighbor node of ⁇ T2, rand1 is ⁇ T2, start , it can be determined that ⁇ T2, start is ⁇ nearest2 ;
  • Step 7 Determine the distance d from ⁇ nearest2 to the nearest obstacle. When it is greater than the preset threshold d min , generate a new node ⁇ T2,new1 according to PSO; otherwise, generate another new node randomly.
  • the new node ⁇ T2,new1 is expanded and grown from the two parts of random learning and target learning, and grows with a certain step in the direction of the new node ⁇ T2,new1 , that is, ⁇ T2, new1 is composed of ( ⁇ T2, rand1 - ⁇ T2, start ) random learning part and ( ⁇ T1, new1 - ⁇ T2, start ) target learning part, the calculation formula is as follows:
  • the value of d min is related to the size of the obstacle
  • ⁇ t is the preset step size step
  • c1 and c2 are the acceleration factors
  • x' goal is the target tree node of the Tree, that is, the target node ⁇ of the current growth of the first exploration tree T1,new1 is the target tree node that is currently closest to the first search tree in the second search tree;
  • Step 8 Carry out collision detection on the new node ⁇ T2, new1 , if it is determined that the new node ⁇ T2, new1 does not collide with obstacles, then add ⁇ T2, new1 to the node set of Tree, otherwise discard the node, and repeat Steps 1 to 4 continue to grow;
  • Step 9 Determine whether the distance between the nearest neighbor nodes of the two exploration trees is less than or equal to the preset step size step. If it is less than or equal to the preset step size step, the two exploration trees can be connected and output a starting position The collision-free path to the target position, otherwise, repeat step 3 to continue searching the growing Tree until the termination condition is met and jump out of the loop, and finally output a collision-free path from the starting position to the target position of the manipulator.
  • this embodiment is based on Particle-swarm Rapid-exploration Random Tree Connect (P-RRT-Connect) for path planning, which has good probability completeness, can avoid falling into local optimum, and can optimize The speed of convergence toward the target not only ensures the effective obstacle avoidance ability, but also improves the convergence rate, so that the robotic arm can successfully avoid obstacles and reach the target quickly.
  • P-RRT-Connect Particle-swarm Rapid-exploration Random Tree Connect
  • Figure 11 is a schematic diagram of a comparison between the path determined based on the path planning method provided by an embodiment of the present application and the path determined based on the path planning method in the related art in a two-dimensional space example, where Fig. 11 shows a schematic diagram of an example under two-dimensional conditions, and the path planning method in the related art takes RRT-Connect as an example.
  • the starting position of the end of the manipulator is set to (0, 0)
  • the target position is set to (100, 100)
  • the environment boundary is x ⁇ [-20, 120], y ⁇ [-20, 120]
  • the obstacle coordinates are (50, 50)
  • the radius is 20.
  • Table 1 The comparison parameters are shown in Table 1 below.
  • Figure 12 is a schematic diagram of the comparison between the path determined based on the path planning method provided by an embodiment of the present application and the path determined based on the path planning method in the related art in a three-dimensional space example, where Figure 12 What is shown is a schematic diagram of an example under a three-dimensional condition, and the path planning method in the related art uses RRT-Connect as an example.
  • the starting position of the end of the manipulator is set to (0, 0, 0)
  • the target position is set to (80, 80, 80)
  • the environment boundary is x ⁇ [-20,100], y ⁇ [- 20, 100], z ⁇ [-20, 100]
  • the obstacle coordinates are (40, 40, 40), and the radius is 30.
  • Table 2 The comparison parameters are shown in Table 2 below.
  • This application implements path planning based on P-RRT-Connect.
  • Two rapidly expanding random trees are grown simultaneously from the starting point and end point of the moving object to search the state space.
  • the newly grown nodes adopt the idea of combining random learning and target learning, which can Accelerate the convergence speed, reduce the path length, and at the same time, by calculating the distance from the newly grown node to the nearest obstacle, it can effectively reduce the number of collisions of the newly grown nodes, reduce the planning time, and finally quickly plan a shorter path for the moving object Optimize the path. Therefore, the path planning method of the present application has:
  • Rationality that is, the determined path can complete the obstacle avoidance task, and the planning result returned by each algorithm can be tracked and executed by the moving object;
  • the returned planning path has better performance than the planning methods of related technologies in terms of time cost, path length and number of created nodes.
  • an embodiment of the present application also provides a path planning device, which includes: a memory, a processor, and a computer program stored in the memory and operable on the processor.
  • the processor and memory can be connected by a bus or other means.
  • the non-transitory software programs and instructions required to realize the path planning methods of the above-mentioned embodiments are stored in the memory, and when executed by the processor, the path planning methods of the above-mentioned embodiments are executed, for example, the above-described execution of the above-described FIG. 1 Method steps S100 to S300, method steps S400 to S500 in FIG. 2 , method steps S210 in FIG. 3 , method steps S220 in FIG. 4 , method steps S230 to S260 in FIG. 5 , method steps S270 in FIG. 6 , Method step S280 in FIG. 7 , method step S290 in FIG. 8 or method step S310 in FIG. 9 .
  • the device embodiments described above are only illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place, or may be distributed to multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
  • an embodiment of the present application also provides a computer-readable storage medium, the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are executed by a processor or a controller, for example, by the above-mentioned Execution by a processor in the device embodiment can cause the processor to execute the path planning method in the above embodiment, for example, execute the method steps S100 to S300 in FIG. 1 and the method steps S400 to S500 in FIG. 2 described above. , method step S210 among Fig. 3, method step S220 among Fig. 4, method step S230 to S260 among Fig. 5, method step S270 among Fig. 6, method step S280 among Fig. 7, method step S290 among Fig. 8 Or the method step S310 in FIG. 9 .
  • the embodiment of the present application includes: obtaining the motion space, starting point position, end point position and obstacle information of the moving object; in the motion space, constructing the first exploration tree from the starting point position, and constructing the second exploration tree from the end point position, wherein , the first search tree and the second search tree grow towards each other according to the obstacle information; when the distance between the first tree node grown by the first search tree and the second tree node grown by the second search tree is not greater than the preset step size, Connect the first search tree and the second search tree to obtain the planned path of the moving object; wherein, the second tree node grows in the following way: from the node set of the second search tree, select the tree node closest to the target sampling point as The third tree node, and select the tree node closest to the second search tree on the first search tree as the target tree node, wherein the target sampling point is obtained by expanding sampling along the second tree node grown last time; when the third tree The shortest distance from the node to the obstacle is not less than the preset distance threshold, and according
  • the path planning space is sought in the way that the first search tree and the second search tree grow oppositely, and the planned path of the moving object is obtained by connecting the first search tree and the second search tree, which can effectively Improve the obstacle avoidance effect of the construction path, and consider the influence of the distance of the relative obstacle, and use the combination of random learning and target learning to build the tree nodes of the exploration tree, which can balance the randomness and targetness of tree node growth, and effectively reduce the generated
  • the possibility of collision between tree nodes and obstacles enables the search tree to converge to the target faster, thereby reducing the path planning time and obtaining a more optimized non-collision continuous path, which can meet the path planning requirements.
  • Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cartridges, tape, magnetic disk storage or other magnetic storage devices, or can Any other medium used to store desired information and which can be accessed by a computer.
  • communication media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

Landscapes

  • Engineering & Computer Science (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Automation & Control Theory (AREA)
  • Robotics (AREA)
  • Mechanical Engineering (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
  • Feedback Control In General (AREA)

Abstract

一种路径规划方法包括：获取移动对象的运动空间、起点位置、终点位置以及障碍物信息(S100)；在运动空间内，从起点位置开始构建第一探索树，从终点位置开始构建第二探索树，第一探索树与第二探索树根据障碍物信息相向生长(S200)；当第一探索树生长的第一树节点与第二探索树生长的第二树节点之间的距离不大于预设步长，连接第一探索树和第二探索树，得到移动对象的规划路径(S300)。这种方法能够有效提升构建路径的避障效果，降低路径规划时间。还涉及一种路径规划装置和计算机可读存储介质。

Description

路径规划方法及装置、可读存储介质
相关申请的交叉引用
本申请基于申请号为202210156637.4、申请日为2022年02月21日的中国专利申请提出，并要求该中国专利申请的优先权，该中国专利申请的全部内容在此引入本申请作为参考。
技术领域
本申请实施例涉及路径导航技术领域，尤其涉及一种路径规划方法及装置、计算机可读存储介质。
背景技术
随着科技的进步，工业移动对象发展迅速，不但可以提高生产效率还可以降低劳动成本，在人类生活中被广泛使用。移动对象诸如机械臂、机器人等的路径规划是工业控制的重要部分，主要指在满足环境约束和移动对象本身要求的情况下，在位姿空间找到一条从起始点到目标点的连续无碰撞路径。目前主流的路径规划算法主要为以快速扩展树算法(Rapid-exploration Random Tree，RRT)为代表的基于采样的规划算法，该算法的随机性强但构建路径用时较长，无法满足当前路径规划要求。
发明内容
以下是对本文详细描述的主题的概述。本概述并非是为了限制权利要求的保护范围。
本申请实施例提供了一种路径规划方法及装置、计算机可读存储介质。
第一方面，本申请实施例提供了一种路径规划方法，包括：获取移动对象的运动空间、起点位置、终点位置以及障碍物信息；在所述运动空间内，从所述起点位置开始构建第一探索树，以及从所述终点位置开始构建第二探索树，其中，所述第一探索树与所述第二探索树根据所述障碍物信息相向生长；当所述第一探索树生长的第一树节点与所述第二探索树生长的第二树节点之间的距离不大于预设步长，连接所述第一探索树和所述第二探索树，得到所述移动对象的规划路径；其中，所述第二树节点按照以下方式生长：从所述第二探索树的节点集合中，选出距目标采样点最近的树节点为第三树节点，以及选出所述第一探索树上距所述第二探索树最近的树节点为目标树节点，其中，所述目标采样点为沿上一次生长的所述第二树节点扩展采样得到；当所述第三树节点到所述障碍物的最近距离不小于预设间距阈值，根据所述第三树节点朝所述目标采样点的方向以及所述第三树节点朝所述目标树节点的方向，生长一个树节点作为所述第二树节点。
第二方面，本申请实施例还提供了一种路径规划装置，包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现如上第一方面所述的路径规划方法。
第三方面，本申请实施例还提供了一种计算机可读存储介质，存储有计算机可执行指令，所述计算机可执行指令用于执行如上第一方面所述的路径规划方法。
本申请的其它特征和优点将在随后的说明书中阐述，并且，部分地从说明书中变得显而易见，或者通过实施本申请而了解。本申请的目的和其他优点可通过在说明书、权利要求书以及附图中所特别指出的结构来实现和获得。
附图说明
附图用来提供对本申请技术方案的进一步理解，并且构成说明书的一部分，与本申请的实施例一起用于解释本申请的技术方案，并不构成对本申请技术方案的限制。
图1是本申请一个实施例提供的路径规划方法的流程图；
图2是本申请一个实施例提供的路径规划方法中，生长第一树节点的流程图；
图3是本申请一个实施例提供的路径规划方法中，构建第一探索树的流程图；
图4是本申请另一个实施例提供的路径规划方法中，构建第一探索树的流程图；
图5是本申请一个实施例提供的路径规划方法中，生长第二树节点的流程图；
图6是本申请另一个实施例提供的路径规划方法中，生长第二树节点的流程图；
图7是本申请一个实施例提供的路径规划方法中，构建第二探索树的流程图；
图8是本申请另一个实施例提供的路径规划方法中，构建第二探索树的流程图；
图9是本申请一个实施例提供的路径规划方法中，得到移动对象的规划路径的流程图；
图10是本申请一个实施例提供的第一探索树和第二探索树的生长扩展示意图；
图11是基于本申请一个实施例提供的路径规划方法所确定的二维路径，与基于相关技术中的路径规划方法所确定的二维路径的对比示意图；
图12是基于本申请一个实施例提供的路径规划方法所确定的三维路径，与基于相关技术中的路径规划方法所确定的三维路径的对比示意图；
图13是本申请一个实施例提供的路径规划装置的示意图。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处所描述的实施例仅用以解释本申请，并不用于限定本申请。
需要注意的是，虽然在装置示意图中进行了功能模块划分，在流程图中示出了逻辑顺序，但是在某些情况下，可以以不同于装置中的模块划分，或流程图中的顺序执行所示出或描述的步骤。说明书和权利要求书及上述附图中的术语“第一”、“第二”等是用于区别类似的对象，而不必用于描述特定的顺序或先后次序。
本申请提供了一种路径规划方法及装置、计算机可读存储介质，采用第一探索树与第二探索树相向生长的方式寻求路径规划空间，并通过连接第一探索树和第二探索树而得到移动对象的规划路径，能够有效提升构建路径的避障效果，并且考虑相对障碍物的距离影响，采用随机学习和目标学习相结合的方式构建探索树的树节点，可以平衡树节点生长的随机性和目标性，有效降低生成的树节点与障碍物的碰撞可能，使得探索树能够更快地向目标收敛，从而降低路径规划时间，得到更加优化的无碰撞连续路径，能够满足路径规划要求。
下面结合附图，对本申请实施例作进一步阐述。
如图1所示，图1是本申请一个实施例提供的路径规划方法的流程图，该路径规划方法 包括但不限于步骤S100至S300。
步骤S100：获取移动对象的运动空间、起点位置、终点位置以及障碍物信息。
在一实施例中，移动对象的起点位置和终点位置即分别为待构建的规划路径的起点和终点，可以为二维、三维甚至更多维下的表示；移动对象的运动空间表征该移动对象的活动空间，用于约束待构建的规划路径的整体范围，使得不会出现超出移动对象的运动空间的情况，可以为二维、三维甚至更多维下的表示；障碍物与移动对象所对应，为移动对象进行移动的阻碍，所设计的规划路径的一个要求即为避开障碍物，因此通过获取障碍物信息可以为生长的树节点的碰撞检测提供依据，以便于设计出符合要求的规划路径。
可以理解地是，移动对象的类型可以为多种，在工业领域中常用的需要进行移动的智能对象均可以为本实施例中的移动对象，比如可以但不限于为机械臂、机器人以及机控车等，移动对象可以在相关应用场景下进行选择设置，这在本实施例中并未限制。
步骤S200：在运动空间内，从起点位置开始构建第一探索树，以及从终点位置开始构建第二探索树，其中，第一探索树与第二探索树根据障碍物信息相向生长；
其中，第二探索树生长的第二树节点可以但不限于按照以下方式生长：
从第二探索树的节点集合中，选出距目标采样点最近的树节点为第三树节点，以及选出第一探索树距第二探索树最近的树节点为目标树节点，其中，目标采样点为沿上一次生长的第二树节点扩展采样得到；
当第三树节点到障碍物的最近距离不小于预设间距阈值，根据第三树节点朝目标采样点的方向以及第三树节点朝目标树节点的方向，生长一个树节点作为第二树节点。
在一实施例中，考虑相对障碍物的距离影响，采用随机学习和目标学习相结合的方式构建探索树的树节点，可以平衡树节点生长的随机性和目标性，有效降低生成的树节点与障碍物的碰撞可能，使得探索树能够更快地向目标收敛，从而降低路径规划时间，得到更加优化的无碰撞连续路径，能够满足路径规划要求。
可以理解地是，第一探索树沿起点位置逐步朝第二探索树拓展生长第一树节点，第二探索树沿终点位置逐步朝第一探索树拓展生长第二树节点，即第一树节点和第二树节点可以为多个且均为逐步扩展得到的，多个第一树节点的集合即为第一探索树的节点集合，多个第二树节点的集合即为第二探索树的节点集合，相应地，目标采样点也可以为多个，每个目标采样点为沿上一次生长的第二树节点扩展采样得到，最终两棵探索树能够不断地相向交替生长而达到收敛，可以看作为一整棵生长树不断地进行生长，直至满足相关条件则可确认生长树生长完毕，进而藉于该生长树就能够得到移动对象的规划路径。
需要说明的是，可以从第一探索树开始生长，也可以从第二探索树开始生长，这对于本申请的各个实施例均适用；当生成第二树节点之后，若需要继续在该第二探索树进行生长，则以当前得到的第二树节点作为生长下一个第二树节点的基础，即可以是循环进行的，例如当前的目标采样点为沿终点位置扩展采样得到，那么类似的，下一次的目标采样点可以基于当前得到的第二树节点扩展采样得到，同样可以基于类似的方式生长得到下一个第二树节点，可以理解地是，生长第一树节点也可以采用循环的方式，不再赘述；通过上述流程所示的循环交替生长，最终能够得到满足要求的第一探索树和第二探索树。
在一实施例中，目标采样点沿终点位置扩展采样的方式可以为多种，本领域技术人员可以根据应用场景进行选择，例如，在无障碍区域，可以基于目标偏置进行随机采样，即根据 均匀概率分布预先生成一个均匀概率P，当预生成的P小于第一预设概率阈值，则目标采样点xrand等于当前目标点，即选出目标树节点为目标采样点xrand，反之，目标采样点xrand在沿上一次生长的第二树节点扩展的运动空间的范围内随机选取，目标采样点xrand的生成公式如下所示：
其中，Uniform(·)为均匀采样函数，P为[0，1]的随机数，gais表示目标偏置概率值，可以但不限于设置为0.05，xgoal为目标解空间，xfree为移动对象的运动空间。
可以理解地是，第二探索树当前生长的目标节点即为第一探索树上距第二探索树当前最近的树节点，因此可以通过第一探索树的生长树节点以确定第二探索树的目标树节点，在后续说明书中将给出实施方式进行详细说明。
在一实施例中，第三树节点到障碍物的最近距离表征相应的最近邻节点与障碍物之间的相对位置关系；当第三树节点到障碍物的最近距离不小于预设间距阈值，则表明最近邻节点位于远离障碍物区域，此时可以通过增大新生长节点的目标学习方面，即增大沿第三树节点朝目标树节点的方向的目标学习部分，同时结合新生长节点的随机学习方面，即保留沿第三树节点朝目标采样点的方向的随机学习部分，使得新生长节点快速地向目标收敛，有利于提高收敛速率，使得移动对象成功避开障碍物快速到达目标。
可以理解地是，预设间距阈值、第一预设概率阈值的取值不限制，可以根据实际应用场景进行设置，例如可以但不限于基于障碍物信息、移动对象的运动空间等进行考虑设置，这在本实施例中并未限制。
在图2的实施方式中，第一探索树生长的第一树节点可以但不限于按照以下方式生长：
步骤S400：从第一探索树的节点集合中，确定距开始采样点最近的树节点为第四树节点，其中，开始采样点为沿上一次生长的第一树节点扩展采样得到；
步骤S500：沿第四树节点朝开始采样点的方向，按照预设步长生长一个树节点作为第一树节点。
在一实施例中，通过采样得到开始采样点，进而从第一探索树的节点集合中，确定开始采样点的最近邻节点为第四树节点，从而能够基于第四树节点朝开始采样点的方向进行随机生长，得到相应的第一树节点，即采用随机生长的方式以确定第一树节点，以使得所构建的规划路径具有一定的概率完备性。
在一实施例中，开始采样点在无障碍物区域可以但不限于按照以下步骤生成：根据均匀概率分布生成开始采样概率；当开始采样概率小于第二预设概率阈值，从第二探索树的节点集合中选出开始树节点为开始采样点，其中，开始树节点为第二探索树上距第一探索树最近的树节点；当开始采样概率大于或等于第二预设概率阈值，沿上一次生长的第一树节点扩展随机选出开始采样点。可以理解地是，开始采样点的获取方式与目标采样点的获取方式类似且原理相同，由于前述实施例已经详细描述了目标采样点的获取方式，因此为免冗余，在此对其不作赘述。
可以理解地是，预设步长为探索树上按顺序生长的两个树节点之间的长度，其数值可以不限制，由本领域技术人员在实际应用场景中进行选择，在一些场景中，当确定预设步长之 后，第一探索树和第二探索树中的所有树节点均可以依据预设步长进行生长。
在图3的实施方式中，步骤S200中的“从起点位置开始构建第一探索树”，包括但不限于步骤S210。
步骤S210：当确定存在第一树节点与障碍物发生碰撞的情况，丢弃第一树节点，生长第二探索树的第二树节点。
在一实施例中，当确定存在第一树节点与障碍物发生碰撞的情况，则说明当前生长的第一树节点不符合路径要求，因此丢弃第一树节点，不将该第一树节点作为第一探索树的一部分，同时转为从另一侧的第二探索树进行第二树节点的生长，从而实现第一探索树与第二探索树的相向交替生长。
在图4的实施方式中，步骤S200中的“从起点位置开始构建第一探索树”，还包括但不限于步骤S220。
步骤S220：当确定存在第一树节点与障碍物未发生碰撞的情况，将第一树节点添加到第一探索树的节点集合中。
在一实施例中，当确定存在第一树节点与障碍物未发生碰撞的情况，则说明当前生长的第一树节点符合路径要求，因此可以保留当前生长的第一树节点并将其添加到第一探索树的节点集合中，以之作为第一探索树的一部分，并能够继续沿着该第一树节点进行第一探索树的生长，直至出现第一树节点与障碍物发生碰撞为止，因此可以实现第一探索树的生长扩展。
在一实施例中，判断第一树节点是否与障碍物发生碰撞的方式可以为多种，例如，可以首先判定生成的第一树节点是否处于障碍物的位置范围之内，若否，则可以进一步判断最近邻接点与生成的第一树节点之间的连线是否与障碍物发生碰撞；若是，则舍弃该生成的第一树节点，换到第二探索树进行生长，否则将该第一树节点添加到第一探索树的节点集合中，成为第一探索树的一部分，即以之作为规划路径的一部分。
在图5的实施方式中，步骤S200中的“根据第三树节点朝目标采样点的方向以及第三树节点朝目标树节点的方向，生长一个树节点作为第二树节点”，包括但不限于步骤S230至S260。
步骤S230：沿第三树节点朝目标采样点的方向，生长一个随机学习向量；
步骤S240：沿第三树节点朝目标树节点的方向，生长一个目标学习向量；
步骤S250：将随机学习向量和目标学习向量相加得到综合学习向量；
步骤S260：沿综合学习向量的方向，按照预设步长生长一个树节点作为第二树节点。
在一实施例中，参考粒子群算法(Particle Swarm Optimization，PSO)，沿第三树节点朝目标采样点的方向生长一个随机学习向量，相当于在沿第三树节点朝目标采样点的方向生长一个随机学习节点，同时沿第三树节点朝目标树节点的方向生长一个目标学习向量，相当于在沿第三树节点朝目标树节点的方向生长一个目标学习节点，即结合随机学习和目标学习的思想以基于第三树节点生长第二树节点，不仅继承了RRT随机性强的优点，还具有PSO收敛速度快的优点，使第二探索树能够更快速的向目标节点生长，减少规划路径长度。
在一实施例中，所生成的第二树节点即考虑了随机学习向量的影响，也考虑了目标学习向量的影响，因此从实际构建路径上呈现为：第三树节点朝第二树节点的方向，位于第三树节点朝目标采样点的方向、第三树节点朝目标树节点的方向之间，从该部分内容可以看出两棵探索树之间的最近邻节点之间的距离会缩短，因此能够减少规划路径长度。
在图6的实施方式中，第二树节点还可以但不限于按照以下方式生长：
步骤S270：当第三树节点到障碍物的最近距离小于预设间距阈值，沿第三树节点朝目标采样点的方向，按照预设步长生长一个树节点作为第二树节点。
在一实施例中，第三树节点到障碍物的最近距离表征相应的最近邻节点与障碍物之间的相对位置关系；当第三树节点到障碍物的最近距离小于预设间距阈值，则表明最近邻节点位于靠近障碍物区域，此时可以通过增大新生长节点的随机学习方面，即增大沿第三树节点朝目标采样点的方向的随机学习部分，使得新生长节点快速地向目标收敛，有利于提高收敛速率，使得移动对象成功避开障碍物快速到达目标。
在图7的实施方式中，步骤S200中的“从终点位置开始构建第二探索树”，包括但不限于步骤S280。
步骤S280：当确定存在第二树节点与障碍物发生碰撞的情况，丢弃第二树节点，生长第一探索树的第一树节点。
在一实施例中，当确定存在第二树节点与障碍物发生碰撞的情况，则说明当前生长的第二树节点不符合路径要求，因此丢弃第二树节点，不将该第二树节点作为第二探索树的一部分，同时转为从另一侧的第一探索树进行第一树节点的生长，从而实现第一探索树与第二探索树的相向交替生长。
在图8的实施方式中，步骤S200中的“从终点位置开始构建第二探索树”，还包括但不限于步骤S290。
步骤S290：当确定存在第二树节点与障碍物未发生碰撞的情况，将第二树节点添加到第二探索树的节点集合中。
在一实施例中，当确定存在第二树节点与障碍物未发生碰撞的情况，则说明当前生长的第二树节点符合路径要求，因此可以保留当前生长的第二树节点，并将其添加到第二探索树的节点集合中，以之作为第二探索树的一部分，并能够继续沿着该第二树节点进行第二探索树的生长，直至出现第二树节点与障碍物发生碰撞为止，因此可以实现第二探索树的生长扩展。
在一实施例中，判断第二树节点是否与障碍物发生碰撞的方式可以为多种，例如，可以首先判定生成的第二树节点是否处于障碍物的位置范围之内，若否，则可以进一步判断最近邻接点与生成的第二树节点之间的连线是否与障碍物发生碰撞，若是则舍弃该生成的第二树节点，换到第一探索树进行生长，否则将该第二树节点添加到第二探索树的节点集合中，成为第二探索树的一部分，即以之作为规划路径的一部分。
步骤S300：当第一探索树生长的第一树节点与第二探索树生长的第二树节点之间的距离不大于预设步长，连接第一探索树和第二探索树，得到移动对象的规划路径。
在一实施例中，采用第一探索树与第二探索树相向生长的方式以寻求路径规划空间，并且通过连接第一探索树和第二探索树而得到移动对象的规划路径，能够有效提升构建路径的避障效果，并且考虑相对障碍物的距离影响，采用随机学习和目标学习相结合的方式构建探索树的树节点，可以平衡树节点生长的随机性和目标性，有效降低生成的树节点与障碍物的碰撞可能，使得探索树能够更快地向目标收敛，从而降低路径规划时间，得到更加优化的无碰撞连续路径，能够满足路径规划要求。
在图9的实施方式中，步骤S300中包括但不限于步骤S310。
步骤S310：按照树节点生长顺序分别连接第一探索树和第二探索树，并连接第一树节点 与第二树节点，得到移动对象的规划路径。
在一实施例中，在确定存在第一树节点与第二树节点之间的距离不大于预设步长的情况，则可以将第一探索树上的按照扩展顺序依次循环生长的各个第一树节点进行连接，以及将第二探索树上的按照扩展顺序依次循环生长的各个第二树节点进行连接，最后将两个探索树的最近节点，即，将距离最近的第一树节点与第二树节点进行连接，从而得到整体上连续的无碰撞规划路径。
以下给出实施方式以说明上述各实施例的工作原理流程。
实施方式一：
如图10所示，图10是本申请一个实施例提供的第一探索树和第二探索树的生长扩展示意图，其中，图10示出的为二维条件下的示意图，其他的三维乃至更多维条件的情况可以依此类推，本实施方式仅以二维的扩展生长情况进行说明
结合图10的实施方式，以机械臂为例，本申请提供的路径规划方法可以按照如下流程执行：
首先，设置待构建路径的问题定义，包括机械臂的运动空间、开始位置、目标位置和障碍物信息。
然后，确定第一探索树T1和第二探索树T2：随机采样可以为在双向快速扩展随机树(Rapid-exploration Random Tree Connect，RRT-Connect)的基础上加入目标偏置部分，并且将机械臂的开始位置作为T1的开始节点χT1,start，并将χT1,start加到T1的节点集合E1中，目标位置xgoal作为T2的开始节点χT2,start，并将χT2,start加到T2的节点集合E2中。
1)生长树Tree为当前正在生长的树，先将Tree设置为T1。
针对T1进行生长的过程，包括：
步骤1、在无碰撞的环境中生成开始采样点，如图10中的χT1,rand1，基于目标偏置进行随机采样，即预先生成一个均匀概率P1，当预生成的P1小于阈值，则开始采样点χT1,rand1等于当前目标点，反之，开始采样点χT1,rand1在沿上一次生长的第一树节点扩展的运动空间的范围内随机选取；
步骤2、如图10所示，在Tree的节点集合中，根据欧氏距离寻找距离χT1,rand1最近的邻节点χnearest1，由于χT1,rand1的最近邻节点为χT1,start，因此可以确定χT1,start为χnearest1，相关的欧式距离的度量函数如下所示：
ρ(x1,x2)＝||x2-x1||；
其中，x1和x2表示Tree中的任意两个节点，ρ(x1,x2)表示x1和x2之间的距离；
步骤3、χnearest1在χT1,rand1方向上以一定的步长生成一个新节点χT1,new1；
步骤4、对新节点χT1,new1进行碰撞检测，若确定新节点χT1,new1未与障碍物发生碰撞，则将χT1,new1添加到Tree的节点集合中，否则舍弃该节点。
2)Tree换到T2进行生长的过程，包括：
步骤5、在无碰撞的环境中生成目标采样点，如图10中的χT2,rand1，基于目标偏置进行随机采样，即预先生成一个均匀概率P2，当预生成的P2小于阈值，则目标采样点χT2,rand1等于当前目标点，反之，目标采样点χT2,rand1在沿上一次生长的第二树节点扩展的运动空间的范围内随机选取；
步骤6、如图10所示，在Tree的节点集合中，根据欧氏距离寻找距离χT2,rand1最近的邻 节点χnearest2，由于χT2,rand1的最近邻节点为χT2,start，因此可以确定χT2,start为χnearest2；
步骤7、判断χnearest2到最近障碍物的距离d，当大于预设阈值dmin时，根据PSO生成新节点χT2,new1，反之则随机生成另一个新节点。
如图10所示，计算χnearest2到最近障碍物的距离d，若小于dmin，则增大随机性，在χT2,rand1方向上以一定的步长生成一个新节点；若大于dmin，则参考PSO的自我学习、社会学习部分的思想，由随机学习和目标学习两部分扩展生长新节点χT2,new1，在新节点χT2,new1方向上以一定的步长生长，即χT2,new1由(χT2,rand1-χT2,start)随机学习部分和(χT1,new1-χT2,start)目标学习部分组成,计算公式如下所示：
其中，dmin的值大小与障碍物的尺寸有关，Δt为预设步长step，c1、c2为加速度因子，x'goal为Tree的目标树节点，即第一探索树当前生长的目标节点χT1,new1，为第二探索树距第一探索树当前最近的目标树节点；
步骤8、对新节点χT2,new1进行碰撞检测，若确定新节点χT2,new1未与障碍物发生碰撞，则将χT2,new1添加到Tree的节点集合中，否则舍弃该节点，再重复步骤1至4继续生长；
步骤9、判断两棵探索树的最近邻节点之间的距离是否小于或等于预设步长step，若小于或等于预设步长step，则可将两棵探索树连接，输出一条从开始位置到目标位置的无碰撞路径，否则重复步骤3继续搜索生长Tree，直到满足终止条件跳出循环，最终输出一条机械臂的从开始位置到目标位置的无碰撞路径。
可以理解地是，除了相比于RRT和改进的RRT-Connect存在优势之外，即使针对常用的以A*算法为代表的基于搜索的规划算法、以遗传算法为代表的基于启发式的规划算法，本实施方式基于粒子群双向快速扩展随机树(Particle-swarm Rapid-exploration Random Tree Connect，P-RRT-Connect)进行路径规划，具有良好的概率完备性，可以避免陷入局部最优，且能够优化朝向目标收敛的速度，不仅保证了具备有效的避障能力，同时也提高了收敛速率，使机械臂成功避开障碍物快速到达目标。
为了更好的证明上述实施例的应用效果，以下给出试验结果进行对比说明。
实施方式二：
如图11所示，图11是在二维空间实例中，基于本申请一个实施例提供的路径规划方法所确定路径，与基于相关技术中的路径规划方法所确定路径的对比示意图，其中，图11示出的为二维条件下的实例示意图，相关技术中的路径规划方法以RRT-Connect为例。
如图11所示，机械臂末端的开始位置设置为(0，0)，目标位置设置为(100，100)，环境边界为x∈[-20，120]，y∈[-20，120]，障碍物坐标为(50，50)，半径为20，对比参数参见如下表1。
表1 P-RRT-Connect和RRT-Connect在二维空间实例的路径相关参数对比表
为了证明本申请采用的P-RRT-Connect的有效可行性，将P-RRT-Connect和RRT-Connect 应用于上述二维空间实例，各自重复运行100次，在同等搜索步长条件下，对算法的运算时间、路径长度和路径节点树的平均值做了对比，从表1中可以看出，本申请的P-RRT-Connect的运算时间、路径长度和路径节点数均比相关技术的RRT-Connect要少，其中，平均的路径规划时间比RRT-Connect约少18％，平均的路径长度比RRT-Connect约少2％，平均的路径节点数比RRT-Connect大约少2％，这直观地说明了P-RRT-Connect的寻优能力、收敛速度都比RRT-Connect好，可行性更高。
实施方式三：
如图12所示，图12是在三维空间实例中，基于本申请一个实施例提供的路径规划方法所确定路径，与基于相关技术中的路径规划方法所确定路径的对比示意图，其中，图12示出的为三维条件下的实例示意图，相关技术中的路径规划方法以RRT-Connect为例。
如图12所示，机械臂末端的开始位置设置为(0，0，0)，目标位置设置为(80，80，80)，环境边界为x∈[-20，100]，y∈[-20，100]，z∈[-20，100]，障碍物坐标为(40，40,40)，半径为30，对比参数参见如下表2。
表2 P-RRT-Connect和RRT-Connect在三维空间实例的路径相关参数对比表
为了证明本申请采用的P-RRT-Connect的有效可行性，将P-RRT-Connect和RRT-Connect应用于上述三维空间实例，各自重复运行多次，在同等搜索步长条件下，对算法的运算时间、路径长度和路径节点树的平均值做了对比，从表2中可以看出，本申请的P-RRT-Connect的运算时间、路径长度和路径节点数均比相关技术的RRT-Connect要少，其中，平均的路径规划时间比RRT-Connect约少25％，平均的路径长度比RRT-Connect约少20％，平均的路径节点数比RRT-Connect大约少12％，这直观地说明了P-RRT-Connect的寻优能力、收敛速度都比RRT-Connect好，可行性更高。
综上所述，从各个实施例和实施方式中可以得出以下结论：
本申请基于P-RRT-Connect实现路径规划，从移动对象的起点位置和终点位置同时生长两棵快速扩展随机树来搜索状态空间，新生长的节点采用随机学习和目标学习相结合的思想，可以加快收敛速度，减少路径长度，同时通过计算新生长的节点到最近障碍物的距离，可以有效的减少新生长的节点的碰撞次数，减少规划时间，最终快速的为移动对象规划出一条较短的优化路径。因此，本申请的路径规划方法具备：
合理性，即所确定的路径能够完成避障任务，每一次算法返回的规划结果可以被移动对象跟踪执行；
概率完备性，若搜索空间存在从起点至终点的无碰撞连续路径，则只要规划时间够长，本申请一定能返回一条合理的规划路径；
优化性，所返回的规划路径在时间花费、路径长度和创建节点数上相比于相关技术的规划方式性能更优。
另外，参照图13，本申请的一个实施例还提供了一种路径规划装置，该路径规划装置包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序。
处理器和存储器可以通过总线或者其他方式连接。
实现上述实施例的路径规划方法所需的非暂态软件程序以及指令存储在存储器中，当被处理器执行时，执行上述各实施例的路径规划方法，例如，执行以上描述的图1中的方法步骤S100至S300、图2中的方法步骤S400至S500、图3中的方法步骤S210、图4中的方法步骤S220、图5中的方法步骤S230至S260、图6中的方法步骤S270、图7中的方法步骤S280、图8中的方法步骤S290或图9中的方法步骤S310。
以上所描述的装置实施例仅仅是示意性的，其中作为分离部件说明的单元可以是或者也可以不是物理上分开的，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部模块来实现本实施例方案的目的。
此外，本申请的一个实施例还提供了一种计算机可读存储介质，该计算机可读存储介质存储有计算机可执行指令，该计算机可执行指令被一个处理器或控制器执行，例如，被上述设备实施例中的一个处理器执行，可使得上述处理器执行上述实施例中的路径规划方法，例如，执行以上描述的图1中的方法步骤S100至S300、图2中的方法步骤S400至S500、图3中的方法步骤S210、图4中的方法步骤S220、图5中的方法步骤S230至S260、图6中的方法步骤S270、图7中的方法步骤S280、图8中的方法步骤S290或图9中的方法步骤S310。
本申请实施例包括：获取移动对象的运动空间、起点位置、终点位置以及障碍物信息；在运动空间内，从起点位置开始构建第一探索树，以及从终点位置开始构建第二探索树，其中，第一探索树与第二探索树根据障碍物信息相向生长；当第一探索树生长的第一树节点与第二探索树生长的第二树节点之间的距离不大于预设步长，连接第一探索树和第二探索树，得到移动对象的规划路径；其中，第二树节点按照以下方式生长：从第二探索树的节点集合中，选出距目标采样点最近的树节点为第三树节点，以及选出第一探索树上距第二探索树最近的树节点为目标树节点，其中，目标采样点为沿上一次生长的第二树节点扩展采样得到；当第三树节点到障碍物的最近距离不小于预设间距阈值，根据第三树节点朝目标采样点的方向以及第三树节点朝目标树节点的方向，生长一个树节点作为第二树节点。根据本申请实施例提供的方案，采用第一探索树与第二探索树相向生长的方式寻求路径规划空间，并通过连接第一探索树和第二探索树而得到移动对象的规划路径，能够有效提升构建路径的避障效果，并且考虑相对障碍物的距离影响，采用随机学习和目标学习相结合的方式构建探索树的树节点，可以平衡树节点生长的随机性和目标性，有效降低生成的树节点与障碍物的碰撞可能，使得探索树能够更快地向目标收敛，从而降低路径规划时间，得到更加优化的无碰撞连续路径，能够满足路径规划要求。
本领域普通技术人员可以理解，上文中所公开方法中的全部或某些步骤、系统可以被实施为软件、固件、硬件及其适当的组合。某些物理组件或所有物理组件可以被实施为由处理器，如中央处理器、数字信号处理器或微处理器执行的软件，或者被实施为硬件，或者被实施为集成电路，如专用集成电路。这样的软件可以分布在计算机可读介质上，计算机可读介质可以包括计算机存储介质(或非暂时性介质)和通信介质(或暂时性介质)。如本领域普通技术人员公知的，术语计算机存储介质包括在用于存储信息(诸如计算机可读指令、数据结构、程序模块或其他数据)的任何方法或技术中实施的易失性和非易失性、可移除和不可移除介质。计算机存储介质包括但不限于RAM、ROM、EEPROM、闪存或其他存储器技术、CD-ROM、数字多功能盘(DVD)或其他光盘存储、磁盒、磁带、磁盘存储或其他磁存储装置、或者可以用于存储期望的信息并且可以被计算机访问的任何其他的介质。此外，本领域普通技术人员 公知的是，通信介质通常包括计算机可读指令、数据结构、程序模块或者诸如载波或其他传输机制之类的调制数据信号中的其他数据，并且可包括任何信息递送介质。
以上是对本申请的若干实施方式进行的说明，但本申请并不局限于上述实施方式，熟悉本领域的技术人员在不违背本申请范围的前提下还可作出种种的等同变形或替换，这些等同的变形或替换均包括在本申请权利要求所限定的范围内。

Claims (13)

  1. 一种路径规划方法，包括：
    获取移动对象的运动空间、起点位置、终点位置以及障碍物信息；
    在所述运动空间内，从所述起点位置开始构建第一探索树，以及从所述终点位置开始构建第二探索树，其中，所述第一探索树与所述第二探索树根据所述障碍物信息相向生长；
    当所述第一探索树生长的第一树节点与所述第二探索树生长的第二树节点之间的距离不大于预设步长，连接所述第一探索树和所述第二探索树，得到所述移动对象的规划路径；
    其中，所述第二树节点按照以下方式生长：
    从所述第二探索树的节点集合中，选出距目标采样点最近的树节点为第三树节点，以及选出所述第一探索树上距所述第二探索树最近的树节点为目标树节点，其中，所述目标采样点为沿上一次生长的所述第二树节点扩展采样得到；
    当所述第三树节点到所述障碍物的最近距离不小于预设间距阈值，根据所述第三树节点朝所述目标采样点的方向以及所述第三树节点朝所述目标树节点的方向，生长一个树节点作为所述第二树节点。
  2. 根据权利要求1所述的路径规划方法，其中，所述根据所述第三树节点朝所述目标采样点的方向以及所述第三树节点朝所述目标树节点的方向，生长一个树节点作为所述第二树节点，包括：
    沿所述第三树节点朝所述目标采样点的方向，生长一个随机学习向量；
    沿所述第三树节点朝所述目标树节点的方向，生长一个目标学习向量；
    将所述随机学习向量和所述目标学习向量相加得到综合学习向量；
    沿所述综合学习向量的方向，按照预设步长生长一个树节点作为所述第二树节点。
  3. 根据权利要求1所述的路径规划方法，其中，所述第二树节点还按照以下方式生长：
    当所述第三树节点到所述障碍物的最近距离小于预设间距阈值，沿所述第三树节点朝所述目标采样点的方向，按照预设步长生长一个树节点作为所述第二树节点。
  4. 根据权利要求1所述的路径规划方法，其中，所述从所述终点位置开始构建第二探索树，包括：
    当确定存在所述第二树节点与所述障碍物发生碰撞的情况，丢弃所述第二树节点，生长所述第一探索树的所述第一树节点。
  5. 根据权利要求4所述的路径规划方法，其中，所述从所述终点位置开始构建第二探索树，还包括：
    当确定存在所述第二树节点与所述障碍物未发生碰撞的情况，将所述第二树节点添加到所述第二探索树的节点集合中。
  6. 根据权利要求1所述的路径规划方法，其中，所述方法还包括在无障碍物区域生成目标采样点的步骤；所述在无障碍物区域生成目标采样点，包括：
    根据均匀概率分布生成目标采样概率；
    当所述目标采样概率小于第一预设概率阈值，选出所述目标树节点为所述目标采样点；
    当所述目标采样概率大于或等于所述第一预设概率阈值，沿上一次生长的所述第二树节点扩展随机选出所述目标采样点。
  7. 根据权利要求1所述的路径规划方法，其中，所述第一树节点按照以下方式生长：
    从所述第一探索树的节点集合中，确定距开始采样点最近的树节点为第四树节点，其中，所述开始采样点为沿上一次生长的所述第一树节点扩展采样得到；
    沿所述第四树节点朝所述开始采样点的方向，按照预设步长生长一个树节点作为所述第一树节点。
  8. 根据权利要求7所述的路径规划方法，其中，所述方法还包括在无障碍物区域生成开始采样点的步骤；所述在无障碍物区域生成开始采样点，包括：
    根据均匀概率分布生成开始采样概率；
    当所述开始采样概率小于第二预设概率阈值，从所述第二探索树的节点集合中选出开始树节点为所述开始采样点，其中，所述开始树节点为所述第二探索树上距所述第一探索树最近的树节点；
    当所述开始采样概率大于或等于所述第二预设概率阈值，沿上一次生长的所述第一树节点扩展随机选出所述开始采样点。
  9. 根据权利要求1所述的路径规划方法，其中，所述从所述起点位置开始构建第一探索树，包括：
    当确定存在所述第一树节点与所述障碍物发生碰撞的情况，丢弃所述第一树节点，生长所述第二探索树的所述第二树节点。
  10. 根据权利要求9所述的路径规划方法，其中，所述从所述起点位置开始构建第一探索树，还包括：
    当确定存在所述第一树节点与所述障碍物未发生碰撞的情况，将所述第一树节点添加到所述第一探索树的节点集合中。
  11. 根据权利要求1至10任意一项所述的路径规划方法，其中，所述连接所述第一探索树和所述第二探索树，得到所述移动对象的规划路径，包括：
    按照树节点生长顺序分别连接所述第一探索树和所述第二探索树，并连接所述第一树节点与所述第二树节点，得到所述移动对象的规划路径。
  12. 一种路径规划装置，包括：存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，所述处理器执行所述计算机程序时实现如权利要求1至11中任意一项所述的路径规划方法。
  13. 一种计算机可读存储介质，存储有计算机可执行指令，所述计算机可执行指令用于执行权利要求1至11中任意一项所述的路径规划方法。
PCT/CN2023/077432 2022-02-21 2023-02-21 路径规划方法及装置、可读存储介质 WO2023155923A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210156637.4 2022-02-21
CN202210156637.4A CN116652927A (zh) 2022-02-21 2022-02-21 路径规划方法及装置、可读存储介质

Publications (1)

Publication Number Publication Date
WO2023155923A1 true WO2023155923A1 (zh) 2023-08-24

Family

ID=87577613

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/077432 WO2023155923A1 (zh) 2022-02-21 2023-02-21 路径规划方法及装置、可读存储介质

Country Status (2)

Country Link
CN (1) CN116652927A (zh)
WO (1) WO2023155923A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117773911A (zh) * 2023-11-03 2024-03-29 广东工业大学 一种复杂环境下工业机器人避障方法
CN118010032A (zh) * 2024-04-09 2024-05-10 中油管道物资装备有限公司 一种运输车路径规划方法、装置及设备
CN118171725A (zh) * 2024-05-13 2024-06-11 南京华康智能科技有限公司 一种改进型机器人路径规划rrt算法

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2009211571A (ja) * 2008-03-06 2009-09-17 Sony Corp 軌道計画装置及び軌道計画方法、並びにコンピュータ・プログラム
US20110106306A1 (en) * 2009-11-02 2011-05-05 Samsung Electronics Co., Ltd. Path planning apparatus of robot and method and computer-readable medium thereof
US20120277907A1 (en) * 2011-04-28 2012-11-01 Waseda University Trajectory planning method, trajectory planning system and trajectory planning and control system
CN109839110A (zh) * 2019-01-09 2019-06-04 浙江大学 一种基于快速随机搜索树的多目标点路径规划方法
CN110509279A (zh) * 2019-09-06 2019-11-29 北京工业大学 一种仿人机械臂的运动路径规划方法及系统
CN112462785A (zh) * 2020-12-04 2021-03-09 厦门大学 一种移动机器人路径规划方法、装置及存储介质
CN112923944A (zh) * 2021-01-29 2021-06-08 的卢技术有限公司 一种自动驾驶路径规划方法、系统及计算机可读存储介质
CN113204238A (zh) * 2021-04-22 2021-08-03 武汉理工大学 一种移动机器人的路径规划方法、设备及存储介质
CN113934206A (zh) * 2021-07-22 2022-01-14 浙江科技学院 移动机器人路径规划方法、装置、计算机设备和存储介质

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2009211571A (ja) * 2008-03-06 2009-09-17 Sony Corp 軌道計画装置及び軌道計画方法、並びにコンピュータ・プログラム
US20110106306A1 (en) * 2009-11-02 2011-05-05 Samsung Electronics Co., Ltd. Path planning apparatus of robot and method and computer-readable medium thereof
US20120277907A1 (en) * 2011-04-28 2012-11-01 Waseda University Trajectory planning method, trajectory planning system and trajectory planning and control system
CN109839110A (zh) * 2019-01-09 2019-06-04 浙江大学 一种基于快速随机搜索树的多目标点路径规划方法
CN110509279A (zh) * 2019-09-06 2019-11-29 北京工业大学 一种仿人机械臂的运动路径规划方法及系统
CN112462785A (zh) * 2020-12-04 2021-03-09 厦门大学 一种移动机器人路径规划方法、装置及存储介质
CN112923944A (zh) * 2021-01-29 2021-06-08 的卢技术有限公司 一种自动驾驶路径规划方法、系统及计算机可读存储介质
CN113204238A (zh) * 2021-04-22 2021-08-03 武汉理工大学 一种移动机器人的路径规划方法、设备及存储介质
CN113934206A (zh) * 2021-07-22 2022-01-14 浙江科技学院 移动机器人路径规划方法、装置、计算机设备和存储介质

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117773911A (zh) * 2023-11-03 2024-03-29 广东工业大学 一种复杂环境下工业机器人避障方法
CN117773911B (zh) * 2023-11-03 2024-05-17 广东工业大学 一种复杂环境下工业机器人避障方法
CN118010032A (zh) * 2024-04-09 2024-05-10 中油管道物资装备有限公司 一种运输车路径规划方法、装置及设备
CN118171725A (zh) * 2024-05-13 2024-06-11 南京华康智能科技有限公司 一种改进型机器人路径规划rrt算法

Also Published As

Publication number Publication date
CN116652927A (zh) 2023-08-29

Similar Documents

Publication Publication Date Title
Adiyatov et al. A novel RRT*-based algorithm for motion planning in dynamic environments
Bency et al. Neural path planning: Fixed time, near-optimal path generation via oracle imitation
CN110703768B (zh) 一种改进型动态rrt*的移动机器人运动规划方法
WO2023155923A1 (zh) 路径规划方法及装置、可读存储介质
JP6711949B2 (ja) １個以上の障害物を回避して始状態から終状態集合まで移動する物体の経路を決定する方法およびシステム
Gan et al. Research on robot motion planning based on RRT algorithm with nonholonomic constraints
Jiang et al. Path planning for robotic manipulator in complex multi-obstacle environment based on improved_RRT
CN108413963A (zh) 基于自学习蚁群算法的条形机器人路径规划方法
CN113858205A (zh) 一种基于改进rrt*的七轴冗余机械臂避障算法
CN112338916A (zh) 基于快速扩展随机树的机械臂避障路径规划方法及系统
CN112947489A (zh) 复杂环境下焊接机器人无碰撞路径规划方法及装置
CN114705196B (zh) 一种用于机器人的自适应启发式全局路径规划方法与系统
An et al. Obstacle avoidance path planning based on improved APF and RRT
KR20170043946A (ko) 이동체의 경로계획 장치 및 그 계획 방법
Zhang et al. An improved path planning algorithm based on RRT
CN115122317A (zh) 基于筛选最优目标构型的冗余机械臂路径规划方法及系统
Al-Ansarry et al. Hybrid RRT-A*: An Improved Path Planning Method for an Autonomous Mobile Robots.
US20210237270A1 (en) Trajectory generation apparatus, multi-link system, and trajectory generation method
Gugan et al. Towards the development of a robust path planner for autonomous drones
Charalampous et al. Real-Time Robot Path Planning for Dynamic Obstacle Avoidance.
CN113741484A (zh) 一种基于概率模型的路径规划方法、系统及介质
Dang et al. A path planning method for indoor robots based on partial & global A-star algorithm
Kroneman et al. A fast two-stage approach for multi-goal path planning in a fruit tree
Zhou et al. A Path Planning Algorithm Based on Improved RRT
Lin et al. Bidirectional homotopy-guided rrt for path planning

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23755923

Country of ref document: EP

Kind code of ref document: A1