WO2021179409A1 - 一种非规则形状移动机器人的路径规划方法 - Google Patents

一种非规则形状移动机器人的路径规划方法 Download PDF

Info

Publication number
WO2021179409A1
WO2021179409A1 PCT/CN2020/086533 CN2020086533W WO2021179409A1 WO 2021179409 A1 WO2021179409 A1 WO 2021179409A1 CN 2020086533 W CN2020086533 W CN 2020086533W WO 2021179409 A1 WO2021179409 A1 WO 2021179409A1
Authority
WO
WIPO (PCT)
Prior art keywords
robot
node
network
path
workspace
Prior art date
Legal status (The legal status 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 status listed.)
Ceased
Application number
PCT/CN2020/086533
Other languages
English (en)
French (fr)
Inventor
赵湛
朱安静
路恩
金明志
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jiangsu University
Original Assignee
Jiangsu University
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 Jiangsu University filed Critical Jiangsu University
Priority to GB2105342.6A priority Critical patent/GB2593058B/en
Publication of WO2021179409A1 publication Critical patent/WO2021179409A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • 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
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • G05D1/0221—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving a learning process
    • 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
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • G05D1/0214—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory in accordance with safety or protection criteria, e.g. avoiding hazardous areas
    • 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
    • G05D1/0088—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots characterized by the autonomous decision making process, e.g. artificial intelligence, predefined behaviours
    • 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
    • G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
    • G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
    • G05D1/0217—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory in accordance with energy consumption, time reduction or distance reduction criteria
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/004—Artificial life, i.e. computing arrangements simulating life
    • G06N3/008—Artificial life, i.e. computing arrangements simulating life based on physical entities controlled by simulated intelligence so as to replicate intelligent life forms, e.g. based on robots replicating pets or humans in their appearance or behaviour
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • G06N3/061—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using biological neurons, e.g. biological neurons connected to an integrated circuit

Definitions

  • the invention belongs to the technical field of mobile robot path planning, and specifically relates to a path planning method for an irregular shape mobile robot.
  • Path planning is a core issue in the research field of mobile robots. Its goal is to find a route that connects the current state and target state of the robot in the work environment, and avoid collisions with obstacles in the environment.
  • the research methods of path planning include cell decomposition method, artificial potential field method and various intelligent algorithms, but these methods are mainly for single-point robots, that is, the robot is regarded as a single element with a regular shape, and the size of the robot is used to affect the working space.
  • the network is divided, and path planning is performed on this basis. In the path planning algorithm, only the robot's 2-degree-of-freedom movement is considered. In reality, the shape of most robots is irregular.
  • the path planning of an irregular shape robot not only needs to consider the movement of 2 degrees of freedom, but also its rotation.
  • the robot is affected by obstacles. It can only move or rotate within a certain angle range. These are important constraints in path planning.
  • the obstacles in the workspace may be smaller than the size of the robot.
  • the current solution is to expand the shape of the obstacles, that is, as long as there are obstacles in the divided workspace grid, the entire grid is regarded as Obstacles, which causes the possible exercise path to be ignored, which is the so-called "path blocked" problem.
  • the safety of the robot must also be considered to avoid “too close” or "too far” problems.
  • the present invention proposes a path planning method for an irregularly shaped mobile robot.
  • the present invention is a planning method that integrates biological neural networks and convolutional neural networks to control path safety to solve complex environments. , The path planning problem of irregular shape mobile robot.
  • This method evenly divides the robot’s workspace into a topological network structure, and performs iterative calculations of bio-neural dynamics, and uses the stable network node function values after iterative calculation to describe the environment information of the workspace; forms a core matrix according to the shape characteristics of the mobile robot, The core matrix and the biological neural network in the workspace are convolved to obtain the movement state matrix and the rotation state matrix of the robot at the network node; using the initial node position and direction of the robot in the workspace as the initial conditions, iteratively search step by step, and build from the initial The node tree from the state to the target state, and finally the node tree is searched backward to obtain the planned path of the robot.
  • both the robot and the target include x, y coordinates and three-dimensional state parameters of the direction ⁇ , taking into account the robot’s 2-degree-of-freedom movement and 1-degree-of-freedom rotational motion, and can control the planning by adjusting the safety factor in the convolution calculation
  • the distance between the path and the obstacle adjusts the safety of the robot's path.
  • the technical solution of the present invention is: a path planning method for an irregularly shaped mobile robot, including the following steps:
  • Construct a workspace network divide the two-dimensional workspace of the robot evenly into a topological workspace network T, and perform biological neural dynamics iterative calculations on the workspace network T to form the biological neural network S of the workspace, and the description of the biological neural network S The environmental information of the entire workspace;
  • Construct the shape core matrix of the robot construct the core matrix according to the shape characteristics of the mobile robot;
  • Construct the movement state matrix and rotation state matrix of the robot perform convolution calculation on the core matrix and the biological neural network in the work space to obtain the movement state matrix and rotation state matrix of the robot at the network node;
  • Build a tree of nodes Take the initial position and direction of the robot in the workspace as initial conditions, and iteratively search step by step to build a tree of nodes from the initial state to the target state;
  • Generate planned path Search the node tree in reverse to obtain the planned path of the robot.
  • the two-dimensional workspace of the robot is evenly divided into a topological workspace network T.
  • Each node of the workspace network T includes free space and obstacles occupying two states.
  • the node function value of the free space is set to 0, and obstacles occupy the node function value is set to 1, the working space of the network node T T I will be adjacent to the eight neighboring nodes, the network workspace T biological neural dynamics iteration, biological neural network constituting the working space S, the biological
  • the function value of the node on the neural network S describes the environmental information of the location of the node, and the biological neural network S describes the environmental information of the entire workspace.
  • a function value X i of the i-th neuron nodes of S i at time t, A, B, D is a non-negative constant parameter set, and there is A>B> D
  • the step of constructing the shape core matrix of the robot is specifically as follows:
  • Each element of the matrix includes two states: free space and robot occupancy.
  • the element value of free space is set to 0, and the robot
  • Extract h 2, ⁇ (x, y), ⁇ 1, 2, ..., 8, to form a 1 ⁇ 8 movement state matrix TSM (x, y), which describes the movement output state of the robot at the current node, that is, whether Able to enter the state of 8 adjacent nodes from the node (x, y);
  • Extract h 1, ⁇ (x, y), ⁇ 1, 2, ..., 8, to form a 1 ⁇ 8 rotation state matrix RSM (x, y), which is composed of '0'or'-1', describing The rotation state of the robot at the node (x, y).
  • the first layer convolution calculation method is:
  • s is the set of subspaces occupied by the biological neural network S where the kernel matrix W ⁇ is located,'*' represents the convolution calculation, and b represents the safety factor, using the symbolic activation function f( ⁇ ), which is defined as
  • the step of constructing the node tree is specifically:
  • search step step step+1, determine whether the target information belongs to the information set P step , if the target information belongs to information Set P step , then stop searching. If the target information does not belong to the information set P step , update the node tree Tree that has been searched, and set the node information set P step to the current state, and perform the next search until the target information is searched until;
  • step 1, 2, ... p, a tree of nodes from the initial state to the target state is formed.
  • the nodes searched by p constitute the path node Path connecting the target position and the initial position, that is, the planned path.
  • the determination of the determined path node is specifically: if there is a horizontally or vertically connected parent node in the parent node set, use it as the walking path node of the robot; The parent node, and use it as the robot walking path node.
  • the actual space described by the nodes of the biological neural network S is equal to the actual space described by the nodes of the core matrix W ⁇ reflecting the shape of the robot.
  • the robot placed in the biological neural network S occupies a core
  • the core matrix W ⁇ reflecting the shape of the robot is placed in the biological neural network S, which can ensure that the node positions of W ⁇ and S correspond one-to-one.
  • the beneficial effect of the present invention is: the present invention integrates biological neural network and convolutional neural network into path planning, and proposes a path planning method for an irregularly shaped mobile robot, which describes the robot's path by constructing a kernel matrix.
  • the shape feature is used for the safe path planning of any irregularly shaped mobile robot in a complex workspace; because the size of the workspace division network is much smaller than the size of the robot, it can describe the characteristics of small obstacles well and solve the dense scattered obstacles The problem of "path blocked" in the working environment; only need to change the safety factor in the convolution calculation, the distance between the planned path and the obstacle can be adjusted, and the safety of the planned path can be easily controlled.
  • FIG. 1 is a schematic diagram of the work space division of a robot according to an embodiment of the present invention.
  • Fig. 2 is a schematic diagram of the connection of neuron nodes in a workspace according to an embodiment of the present invention.
  • Fig. 3 is a schematic diagram of the construction of a biological neural network in a workspace according to an embodiment of the present invention.
  • Fig. 4 is a schematic diagram of the construction of a shape core matrix of a robot according to an embodiment of the present invention.
  • FIG. 5 is a schematic diagram of calculation of a convolutional neural network according to an embodiment of the present invention.
  • FIG. 6 is a schematic diagram of a secondary convolution calculation kernel matrix according to an embodiment of the present invention.
  • Fig. 7 is a flow chart of constructing a node tree according to an embodiment of the present invention.
  • Fig. 8 is a flowchart of generating a planned route according to an embodiment of the present invention.
  • Fig. 9 is a plan path diagram in a U-shaped obstacle working environment according to an embodiment of the present invention.
  • Fig. 10 is a plan path diagram in a densely scattered obstacle work environment according to an embodiment of the present invention.
  • a path planning method for an irregularly shaped mobile robot includes the following steps:
  • Construct a workspace network divide the two-dimensional workspace of the robot evenly into a topological workspace network T, and perform biological neural dynamics iterative calculations on the workspace network T to form the biological neural network S of the workspace, and the description of the biological neural network S The environmental information of the entire workspace;
  • Construct the shape core matrix of the robot construct the core matrix according to the shape characteristics of the mobile robot;
  • Construct the movement state matrix and rotation state matrix of the robot perform convolution calculation on the core matrix and the biological neural network in the work space to obtain the movement state matrix and rotation state matrix of the robot at the network node;
  • Build a tree of nodes Take the initial position and direction of the robot in the workspace as initial conditions, and iteratively search step by step to build a tree of nodes from the initial state to the target state;
  • the steps of constructing a workspace network are specifically as follows:
  • the two-dimensional workspace of the mobile robot is evenly divided into a topological workspace network T along the x and y directions.
  • it is divided into a 100 ⁇ 100 network structure.
  • Each node T of the workspace network T (x, y) includes two states: free space and obstacle occupancy.
  • the free space is shown in the white area in Figure 1.
  • the obstacle occupancy is shown in the figure
  • the work space network T is subjected to the iterative calculation of biological neural dynamics to form the biological neural network S of the work space.
  • the calculation formula is
  • a function value X i of the i-th neuron nodes of S i at time t, A, B, D is a non-negative constant parameter set, and there is A>B> D
  • I i is a biological neural the network node S i external stimulus
  • I i E
  • I i -E
  • I j is the j-th adjacent neuron
  • j 1, 2 ...8
  • the connection weight of adjacent neurons w ij 1/d
  • d is the distance between the i-th and j-th neuron nodes
  • the function value of the node on the biological neural network S describes the environmental information of the location of the node, and the biological neural network S describes the environmental information of the entire workspace.
  • the steps of constructing the shape core matrix of the robot are specifically as follows:
  • the shape of the mobile robot can be arbitrary;
  • the element W(m,n) includes two states: free space and robot occupation.
  • the actual space described by the nodes of the constructed biological neural network S should be equal to the actual space described by the nodes of the robot shape core matrix W ⁇ .
  • the robot placed in the biological neural network S occupies a sub-network equal in size to the core matrix W.
  • the calculation structure of the convolutional neural network is shown in Figure 5.
  • the 8 groups of core matrices W ⁇ reflecting the shape of the robot are convolved with the biological neural network S to obtain 8 one-layer convolutional networks H 1, ⁇ , convolution calculation
  • the method is
  • s is the set of subspaces occupied by the biological neural network S where the kernel matrix W ⁇ is located, S p and q are the subspaces of s,'*' represents the convolution calculation, b represents the safety factor, and the symbolic activation function f is used.
  • Use M ⁇ to perform the secondary convolution calculation on the one-layer convolutional network H 1, ⁇ respectively, and obtain 8 two-layer convolutional networks H 2, ⁇ .
  • the safety factor b is set to 0 in the convolution calculation.
  • Extract h 2, ⁇ (x, y), ⁇ 1, 2, ..., 8, to form a 1 ⁇ 8 movement state matrix TSM (x, y), which is composed of '0'or'-1', describing The robot's mobile output state at the current node, that is, whether it can go from node (x, y) to the state of 8 adjacent nodes.
  • TSM(x,y) [0,-1,-1,0,0,0,-1,0], which means that the robot can travel from the node (x,y) to the adjacent 5 nodes (x+ 1, y), (x+1, y-1), (x, y-1), (x-1, y-1) and (x-1, y+1).
  • Extract h 1, ⁇ (x, y), ⁇ 1, 2, ..., 8, to form a 1 ⁇ 8 rotation state matrix RSM (x, y), which is composed of '0'or'-1', describing The rotation state of the robot at the node (x, y).
  • the steps of constructing a node tree are specifically:
  • the node information set P step includes the position and direction of the robot. Only when they are equal to the target position (x p , y p ) and the direction ⁇ p , the robot reaches the target.
  • step 1, 2, ... p, a tree of nodes from the initial state to the target state is formed.
  • the program After the search is completed, the program returns to the constructed search node tree Tree.
  • search for the parent node connected to it to form a parent node set if there is a parent node connected horizontally or vertically in the parent node set, use it as the robot walking path node, if not, look for the parent node
  • the parent node with a smaller angle of rotation in the current direction of the robot is used as the node of the robot's walking path.
  • the program returns the path node Path that connects the target location and the initial location, that is, the planned path.
  • the distance between the path and the obstacle can be controlled, and as b increases, the distance between the path and the obstacle The distance increases to ensure the safety of the robot.
  • the planned path is the shortest path under a certain safety factor b.
  • the arrow in Figure 9 is the forward direction.
  • the path planning of the irregularly shaped mobile robot takes into account the robot’s 3-degree-of-freedom motion in the workspace, that is, 2-degree-of-freedom translation and 1-degree-of-freedom rotation.
  • the robot needs to meet the requirements at the same time. Only when the target position (x p , y p ) and direction ⁇ p are reached can the path be generated.
  • the planned path can be changed by setting the safety factor b.
  • b3>b2>b1 it can be seen that as the safety factor increases, the planned path will also change accordingly.
  • the planned path and obstacles The distance increases accordingly to ensure the safety of robot movement.
  • the actual space described by the nodes of the constructed biological neural network S is equal to the actual space described by the nodes of the core matrix W ⁇ of the robot shape.
  • the robot occupies a sub-network in S.
  • W ⁇ has a one-to-one correspondence with the node positions of S.
  • Figure 10 shows the results of the planned path in the working environment of densely scattered obstacles. Set the safety factor b, b2>b1, and you can also adjust the distance between the path and the obstacle.
  • the present invention is used for safe path planning of any irregularly shaped mobile robot in a complex workspace.
  • the workspace of the robot is evenly divided into a topological network structure, and the iterative calculation of biological neural dynamics is carried out, and iterative calculation is used to stabilize the
  • the network node function value describes the environment information of the working space;
  • the core matrix is formed according to the shape characteristics of the mobile robot, and the core matrix is convolved with the biological neural network of the working space to obtain the movement state matrix and the rotation state matrix of the robot at the network node;
  • iteratively search step by step construct a node tree from the initial state to the target state, and finally search the node tree backwards to obtain the robot's planned path.
  • the robot can be of any irregular shape, taking into account the robot's 3-degree-of-freedom motion in the plane space, and adding a safety factor factor in the convolution calculation, controlling the distance between the planned path and the obstacle, and solving the dense dispersion.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Automation & Control Theory (AREA)
  • Remote Sensing (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Business, Economics & Management (AREA)
  • Game Theory and Decision Science (AREA)
  • Medical Informatics (AREA)
  • Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
  • Feedback Control In General (AREA)
  • Manipulator (AREA)

Abstract

一种非规则形状移动机器人的路径规划方法，包括以下步骤：将机器人的二维工作空间均匀划分成拓扑的工作空间网络T，并进行生物神经动力学迭代计算，构成工作空间的生物神经网络S；根据移动机器人的形状特征构成核矩阵；将核矩阵与工作空间的生物神经网络进行卷积计算，获得机器人在网络节点的移动状态矩阵和转动状态矩阵；以机器人在工作空间初始位置和方向为初始条件，逐步迭代搜索，构建从初始状态到目标状态的节点树Tree；反向搜索节点树Tree，获得机器人的规划路径。机器人可以是任意非规则形状，考虑了机器人的在平面空间的3自由度运动，并且在卷积计算中增加安全系数因子，控制规划路径与障碍物之间的距离，调节行使路径的安全性。

Description

一种非规则形状移动机器人的路径规划方法 技术领域
本发明属于移动机器人路径规划技术领域，具体涉及一种非规则形状移动机器人的路径规划方法。
背景技术
路径规划是移动机器人研究领域的一个核心问题，它的目标就是在工作环境中寻找一条连接机器人当前状态和目标状态的行使路线，并且避免与环境中的障碍物发生碰撞。目前，路径规划的研究方法包括细胞分解法、人工势场法以及各类智能算法等，但这些方法主要是针对单点机器人，即将机器人视为形状规则的单个元素，以机器人的尺寸对工作空间进行网络划分，在此基础上进行路径规划，路径规划算法中只考虑机器人的2自由度移动。实际情况下，大多数机器人的形状是非规则的，非规则形状机器人的路径规划不仅需要考虑2自由度移动，还需要考虑其转动，例如，在某些复杂环境下，受到障碍物的影响，机器人只能作移动运动或在一定角度范围内转动，这些是路径规划中的重要约束。其次，工作空间中的障碍物有可能比机器人的尺寸要小，现在的解决办法是将障碍物形状进行拓展，即划分的工作空间网格内只要有障碍物，就将整个网格均视为障碍物，这就造成可能的行使路径被忽略了，即所谓的“path blocked”问题。此外，路径规划中，还需要考虑机器人行使的安全性，避免“too close”或者“too far”问题。
发明内容
针对上述技术问题，本发明提出了一种非规则形状移动机器人的路径规划方法，本发明是将生物神经网络和卷积神经网络相融合，可以控制路径安全性的规划方法，以解决复杂环境下、非规则形状移动机器人的路径规划难题。该方法将机器人的工作空间均匀划分成拓扑网络结构，并进行生物神经动力学迭代计算，采用迭代计算稳定后的网络节点函数值描述工作空间的环境信息；根据移动机器人的形状特征构成核矩阵，将核矩阵与工作空间的生物神经网络进行卷积计算，获得机器人在网络节点的移动状态矩阵和转动状态矩阵；以机器人在工作空间初始节点位置和方向为初始条件，逐步迭代搜索，构建从初始状态到目标状态的节点树，最后反向搜索节点树，获得机器人的规划路径。
路径规划中，机器人和目标均包括x、y坐标和方向θ三维状态参数，考虑了机器人的2自由度移动和1自由度转动运动，并且可以通过调节卷积计算中的安全系数因子，控制规划 路径与障碍物之间的距离，调节机器人行使路径的安全性。
本发明的技术方案是：一种非规则形状移动机器人的路径规划方法，包括以下步骤：
构建工作空间网络：将机器人的二维工作空间均匀划分成拓扑的工作空间网络T，并对工作空间网络T进行生物神经动力学迭代计算，构成工作空间的生物神经网络S，生物神经网络S描述了整个工作空间的环境信息；
构建机器人形状核矩阵：根据移动机器人的形状特征构成核矩阵；
构建机器人的移动状态矩阵和转动状态矩阵：将核矩阵与工作空间的生物神经网络进行卷积计算，获得机器人在网络节点的移动状态矩阵和转动状态矩阵；
构建节点树Tree：以机器人在工作空间初始位置和方向为初始条件，逐步迭代搜索，构建从初始状态到目标状态的节点树Tree；
生成规划路径：反向搜索节点树Tree，获得机器人的规划路径。
上述方案中，所述构建工作空间网络的步骤具体为：
将机器人的二维工作空间均匀划分成拓扑的工作空间网络T，工作空间网络T的每个节点包括自由空间和障碍物占据两个状态，自由空间的节点函数值设定为0，障碍物占据的节点函数值设定为1，工作空间网络T中的节点T i会与周边8个节点相邻，将工作空间网络T进行生物神经动力学迭代计算，构成工作空间的生物神经网络S，生物神经网络S上节点的函数值描述了该节点所在位置的环境信息，生物神经网络S则描述了整个工作空间的环境信息。
上述方案中，所述生物神经网络S迭代计算公式为：
Figure PCTCN2020086533-appb-000001
公式一中，x i为第i个神经元节点S i的在t时刻的函数值，A、B、D为设定的非负恒定参数，且有A>B>D，I i为生物神经网络节点S i的外界激励，且有
Figure PCTCN2020086533-appb-000002
障碍物占据节点的外界激励为I i＝E，自由空间节点的外界激励为I i＝-E，且有E>>B，I j为第j个相邻的神经元，j＝1，2…8，公式一迭代算计法中区分自由空间和障碍物的阈值函数定义为：[a] +＝max{a,0}、[a] -＝max{-a,0}，相邻神经元的连接权重w ij＝1/d，d为第i和第j个神经元节点之间的距离，k＝8为周边8个相邻的神经元。
上述方案中，所述构建机器人形状核矩阵步骤具体为：
将移动机器人所占据区域扩展成规则矩形或正方形区域，并进行均匀划分，构成核矩阵W，矩阵的每个元素包括自由空间和机器人占据两个状态，自由空间的元素值设定为0，机器人占据的元素值设定为1，将机器人在工作空间内，间隔45°进行旋转，构建反映机器人8 个方向的核矩阵W θ，θ＝1，2，…，8。
进一步的，所述构建机器人的移动状态矩阵和转动状态矩阵的步骤具体为：
将8组反映机器人形状的核矩阵W θ分别与生物神经网络S进行卷积计算，得到8个一层卷积网络H 1,θ，H 1,θ网络中每个节点的输出h 1,θ(x，y)表示机器人是否能够以θ方向放置于工作空间节点(x，y)，且h 1,θ(x，y)＝0表示能够放置，h 1,θ(x，y)＝-1表示不能够放置；
构建8个3×3的核矩阵M θ，用M θ分别对一层卷积网络H 1,θ进行二次卷积计算，得到8个二层卷积网络H 2,θ，H 2,θ网络中每个节点的输出h 2,θ(x，y)表示机器人是否能够以θ方向从当前节点(x，y)行使进入相邻的节点，且h 2,θ(x，y)＝0表示能够行驶进入，h 2,θ(x，y)＝-1表示不能够行驶进入；
提取h 2,θ(x，y)，θ＝1，2，…，8，构成一个1×8的移动状态矩阵TSM(x，y)，描述了机器人在当前节点的移动输出状态，即是否能够从节点(x，y)行使进入相邻的8个节点的状态；
提取h 1,θ(x，y)，θ＝1，2，…，8，构成一个1×8的转动状态矩阵RSM(x，y)，它由‘0’或‘-1’组成，描述机器人在节点(x，y)的转动状态。
进一步的，所述第一层卷积计算方法为：
Figure PCTCN2020086533-appb-000003
公式二中，s为核矩阵W θ所在生物神经网络S中所占据的子空间集合，‘*’代表卷积计算，b表示安全系数，采用符号激活函数f(·)，定义为
Figure PCTCN2020086533-appb-000004
上述方案中，所述构建节点树的步骤具体为：
以机器人的初始状态，搜索机器人能够进入的相邻神经元信息，构成一组节点信息集合P step，搜索步长step＝step+1，判断目标信息是否属于信息集合P step，如果目标信息属于信息集合P step，则停止搜索，如果目标信息不属于信息集合P step，则更新已经搜索的节点树Tree，并将节点信息集合P step设定为当前状态，进行下一步搜索，直到搜索到目标信息为止；
由每一步长的节点集合P step，step＝1，2，…p，构成一个从初始状态到目标状态的节点树Tree。
上述方案中，所述生成规划路径具体步骤为：
根据所述节点树Tree，输入机器人的初始位置(x 0，y 0)和方向θ 0，输入目标位置(x p，y p)和方向θ p；
从目标信息节点出发，搜索与之相连的父节点，构成父节点集合，确定路径节点，更新 路径节点Path，然后进行循环搜索，直到步长step＝1，由每一步长step＝1，2，…p搜索到的节点构成连接目标位置和初始位置的路径节点Path，即规划路径。
进一步的，所述确定路径节点的确定具体为：如果父节点集合中有水平或垂直相连的父节点，则将其作为机器人行走路径节点，如果没有，则寻找与机器人当前方向转动角度较小的父节点，并将其作为机器人行走路径节点。
上述方案中，所述生物神经网络S的节点所描述的实际空间与反映机器人形状的核矩阵W θ的节点所描述的实际空间相等，机器人放置于生物神经网络S中所占据的是一个与核矩阵W大小相等的子网络，将反映机器人形状的核矩阵W θ放置于生物神经网络S中，能够保证W θ与S的节点位置一一对应。
与现有技术相比，本发明的有益效果是：本发明将生物神经网络和卷积神经网络融入路径规划中，提出一种非规则形状移动机器人的路径规划方法，通过构建核矩阵描述机器人的形状特征，用于任意非规则形状的移动机器人在复杂工作空间下的安全路径规划；由于工作空间划分网络的尺寸远小于机器人的尺寸，可以很好的描述小障碍物特征，解决密集散点障碍工作环境下“path blocked”难题；只需要改变卷积计算中安全系数，就可实现规划路径与障碍物的距离的调节，方便地控制规划路径的安全性。
附图说明
图1是本发明一实施方式的机器人的工作空间划分示意图。
图2是本发明一实施方式的工作空间神经元节点连接示意图。
图3是本发明一实施方式的工作空间的生物神经网络构建示意图。
图4是本发明一实施方式的机器人形状核矩阵构建示意图。
图5是本发明一实施方式的卷积神经网络计算示意图。
图6是本发明一实施方式的二次卷积计算核矩阵示意图。
图7是本发明一实施方式的节点树的构建流程图。
图8是本发明一实施方式的生成规划路径的流程图。
图9是本发明一实施方式的在U型障碍工作环境下的规划路径图。
图10是本发明一实施方式的在密集散点障碍工作环境下的规划路径图。
具体实施方式
下面详细描述本发明的实施例，所述实施例的示例在附图中示出，其中自始至终相同或类似的标号表示相同或类似的元件或具有相同或类似功能的元件。下面通过参考附图描述的实施例是示例性的，旨在用于解释本发明，而不能理解为对本发明的限制。
一种非规则形状移动机器人的路径规划方法，包括以下步骤：
构建工作空间网络：将机器人的二维工作空间均匀划分成拓扑的工作空间网络T，并对工作空间网络T进行生物神经动力学迭代计算，构成工作空间的生物神经网络S，生物神经网络S描述了整个工作空间的环境信息；
构建机器人形状核矩阵：根据移动机器人的形状特征构成核矩阵；
构建机器人的移动状态矩阵和转动状态矩阵：将核矩阵与工作空间的生物神经网络进行卷积计算，获得机器人在网络节点的移动状态矩阵和转动状态矩阵；
构建节点树Tree：以机器人在工作空间初始位置和方向为初始条件，逐步迭代搜索，构建从初始状态到目标状态的节点树Tree；
生成规划路径：最后反向搜索节点树Tree，获得机器人的规划路径。
所述构建工作空间网络的步骤具体为：
如图1所示，将移动机器人的二维工作空间沿x和y方向均匀划分成拓扑的工作空间网络T，图1中划分成100×100的网络结构，工作空间网络T的每个节点T(x，y)包括自由空间和障碍物占据两个状态，自由空间如图1中白色区域所示，自由空间的节点函数值设定为T(x，y)＝0，障碍物占据如图1中黑色区域所示，障碍物占据的节点函数值设定为T(x，y)＝1。
如图2所示，工作空间网络T中的节点T i的坐标为(x，y)，节点T i会与周边8个节点相邻T j，j＝1，2，…，8。
将工作空间网络T进行生物神经动力学迭代计算，构成工作空间的生物神经网络S，计算公式为
Figure PCTCN2020086533-appb-000005
公式一中，x i为第i个神经元节点S i的在t时刻的函数值，A、B、D为设定的非负恒定参数，且有A>B>D，I i为生物神经网络节点S i的外界激励，且有
Figure PCTCN2020086533-appb-000006
障碍物占据节点的外界激励为I i＝E，自由空间节点的外界激励为I i＝-E，且有E>>B，I j为第j个相邻的神经元，j＝1，2…8，式中的阈值函数定义为：[a] +＝max{a,0}、[a] -＝max{-a,0}，相邻神经元的连接权重w ij＝1/d，d为第i和第j个神经元节点之间的距离，k＝8为周边8个相邻的神经元。根据E>>B，A>B，B>D设定参数A、B、D、E，可以使生物神经网络S中，障碍物占据节点的函数值远大于自由空间节点的函数值，两者之间形成连续的梯度过渡，使得自由空间节点的函数值随着与障碍物距离的增大而逐步减小。
当设定参数A＝10、B＝1、D＝0、E＝50时，迭代计算稳定后得到的工作空间生物神经网络S如图3所示。
生物神经网络S上节点的函数值描述了该节点所在位置的环境信息，生物神经网络S则描述了整个工作空间的环境信息。
所述构建机器人形状核矩阵的步骤具体为：
移动机器人的形状可以是任意的；
假设为“箭头”形状移动机器人，将移动机器人所占据区域扩展成规则的矩形或正方形区域，并进行均匀划分，构成核矩阵W，图4中划分为9×9的正方形矩阵，矩阵的每个元素W(m，n)包括自由空间和机器人占据两个状态，自由空间的元素值设定为W(m，n)＝0，机器人占据的元素值设定为W(m，n)＝1。
将机器人在工作空间内，间隔45°进行旋转，构建反映机器人8个方向的核矩阵W θ，θ表示机器人前进方向，θ＝1，2，…，8，如图4所示。
构建的生物神经网络S的节点所描述的实际空间应该与机器人形状核矩阵W θ的节点所描述的实际空间相等。
机器人放置于生物神经网络S中所占据的是一个与核矩阵W大小相等的子网络。
将机器人形状核矩阵W θ放置于生物神经网络S中，能够保证W θ与S的节点位置一一对应。
所述构建机器人的移动状态矩阵和转动状态矩阵的步骤具体为：
卷积神经网络计算结构参阅图5，构建的核矩阵参阅图6。
卷积神经网络计算结构如图5所示，将8组反映机器人形状的核矩阵W θ分别与生物神经网络S进行卷积计算，得到8个一层卷积网络H 1,θ，卷积计算方法为
Figure PCTCN2020086533-appb-000007
公式中，s为核矩阵W θ所在生物神经网络S中所占据的子空间集合，S p,q为s的子空间，‘*’代表卷积计算，b表示安全系数，采用符号激活函数f(·)，定
Figure PCTCN2020086533-appb-000008
卷积计算后，H 1,θ网络中每个节点的输出h 1,θ(x，y)表示机器人是否能够以θ方向放置于工作空间节点(x，y)，且h 1,θ(x，y)＝0表示能够放置，h 1,θ(x，y)＝-1表示不能够放置。
如图6所示，构建8个3×3的核矩阵M θ，θ＝1，2，…，8。
用M θ分别对一层卷积网络H 1,θ进行二次卷积计算，得到8个二层卷积网络H 2,θ，卷积计算中安全系数b设定为0。
二次卷积H 2,θ网络中每个节点的输出h 2,θ(x，y)表示机器人是否能够以θ方向从当前节 点(x，y)行使进入相邻的节点，且h 2,θ(x，y)＝0表示能够行驶进入，h 2,θ(x，y)＝-1表示不能够行驶进入。
提取h 2,θ(x，y)，θ＝1，2，…，8，构成一个1×8的移动状态矩阵TSM(x，y)，它由‘0’或‘-1’组成，描述了机器人在当前节点的移动输出状态，即是否能够从节点(x，y)行使进入相邻的8个节点的状态。
例如，TSM(x，y)＝[0,-1,-1,0,0,0,-1,0]，表示机器人可以从节点(x，y)行使进入相邻5个节点(x+1，y)、(x+1，y-1)、(x，y-1)、(x-1，y-1)和(x-1，y+1)。
提取h 1,θ(x，y)，θ＝1，2，…，8，构成一个1×8的转动状态矩阵RSM(x，y)，它由‘0’或‘-1’组成，描述机器人在节点(x，y)的转动状态。
例如，RSM(x，y)＝[-1,-1,-1,0,0,0,-1,0]，表示机器人在节点(i，j)位置，可以在θ＝4，5，和6范围内转动。
所述构建节点树的步骤具体为：
如图7所示，首先进行参数初始化：构建机器人工作空间信息，构建工作空间网络T；根据移动机器人的形状构建8个核矩阵W θ；输入机器人的初始位置(x 0，y 0)和方向θ 0，输入目标位置(x p，y p)和方向θ p；设定规划路径的安全系数b；步长step＝0。
其次，根据工作空间网络T计算工作空间的生物神经网络S；通过卷积计算，获得生物神经网络S上机器人在每个节点的移动状态矩阵TSM和转动状态矩阵RSM。
然后，进入循环搜索程序。
以机器人的初始状态，搜索机器人能够进入的相邻神经元信息，构成一组节点信息集合P step，搜索步长step＝step+1；
判断目标信息是否属于信息集合P step，如果目标信息属于信息集合P step，则停止搜索；如果目标信息不属于信息集合P step，则更新已经搜索的节点树Tree，并将节点信息集合P step设定为当前状态，进行下一步搜索，直到搜索到目标信息为止。
节点信息集合P step包括机器人的位置和方向，只有当它们与目标位置(x p，y p)和方向θ p均相等时，才说明机器人到达目标。
由每一步长的节点集合P step，step＝1，2，…p，构成一个从初始状态到目标状态的节点树Tree。
搜索完成后，程序返回构建的搜索节点树Tree。
所述生成规划路径具体步骤为：
如图8所示：首先，输入已经建立的搜索树Tree，输入机器人的初始位置(x 0，y 0)和方向θ 0，输入目标位置(x p，y p)和方向θ p；步长step＝step-1；
其次，从目标信息节点出发，搜索与之相连的父节点，构成父节点集合；如果父节点集合中有水平或垂直相连的父节点，则将其作为机器人行走路径节点，如果没有，则寻找与机器人当前方向转动角度较小的父节点，并将其作为机器人行走路径节点。
将上一步确定的路径节点作为当前节点，更新路径节点Path，然后进行循环搜索，直到步长step＝1。
由每一步长step＝1，2，…p搜索到的节点增加到路径节点Path，即逐步更新规划路径Path。
程序返回连接目标位置和初始位置的路径节点Path，即规划路径。
节点树Tree的构建过程中，只需要通过设定第一层卷积计算中的安全系数b，可以控制路径与障碍物之间的距离，且随着b的增大，路径与障碍物之间的距离随之增加，以保证机器人行使的安全性。
规划得到的路径是在一定安全系数b情况下的最短路径。
U型障碍工作环境下的规划路径：
如图9所示，图9中箭头为前进方向，非规则形状移动机器人的路径规划考虑了机器人在工作空间内的3自由度运动，即2自由度平移和1自由度转动，机器人需要同时满足到达目标位置(x p，y p)和方向θ p，才能生成路径。
可以通过设定安全系数b，改变所规划的路径，图9中，b3>b2>b1，可见随着安全系数的增大，所规划的路径也会随之作改变，规划的路径与障碍物的距离随之增大，以保证机器人运动的安全性。
密集散点障碍工作环境下的规划路：
如图10所示，本发明中，构建的生物神经网络S的节点所描述的实际空间与机器人形状的核矩阵W θ的节点所描述的实际空间相等，机器人在S中是占据一个子网络，且W θ与S的节点位置一一对应。这就使得生物神经网络S的节点所描述的实际空间要远小于机器人的尺寸，它可以更好的描述小障碍环境，图10给出了在密集散点障碍工作环境下的规划路径结果，通过设定安全系数b，b2>b1，也可以调节路径与障碍物的距离。
可见本发明是用于任意非规则形状的移动机器人在复杂工作空间下的安全路径规划，将机器人的工作空间均匀划分成拓扑网络结构，并进行生物神经动力学迭代计算，采用迭代计算稳定后的网络节点函数值描述工作空间的环境信息；根据移动机器人的形状特征构成核矩阵，将核矩阵与工作空间的生物神经网络进行卷积计算，获得机器人在网络节点的移动状态矩阵和转动状态矩阵；以机器人在工作空间初始位置和方向为初始条件，逐步迭代搜索，构建从初始状态到目标状态的节点树，最后反向搜索节点树，获得机器人的规划路径。该方法 中，机器人可以是任意非规则形状，考虑了机器人的在平面空间的3自由度运动，并且在卷积计算中增加安全系数因子，控制规划路径与障碍物之间的距离，解决密集散点障碍工作环境下“path blocked”难题；只需要改变卷积计算中安全系数，就可实现规划路径与障碍物的距离的调节，方便地控制规划路径的安全性。
上文所列出的一系列的详细说明仅仅是针对本发明的可行性实施例的具体说明，它们并非用以限制本发明的保护范围，凡未脱离本发明技艺精神所作的等效实施例或变更均应包含在本发明的保护范围之内。

Claims (10)

  1. 一种非规则形状移动机器人的路径规划方法，其特征在于，包括以下步骤：
    构建工作空间网络：将机器人的二维工作空间均匀划分成拓扑的工作空间网络T，并对工作空间网络T进行生物神经动力学迭代计算，构成工作空间的生物神经网络S，生物神经网络S描述了整个工作空间的环境信息；
    构建机器人形状核矩阵：根据移动机器人的形状特征构成核矩阵；
    构建机器人的移动状态矩阵和转动状态矩阵：将核矩阵与工作空间的生物神经网络进行卷积计算，获得机器人在网络节点的移动状态矩阵和转动状态矩阵；
    构建节点树Tree：以机器人在工作空间初始位置和方向为初始条件，逐步迭代搜索，构建从初始状态到目标状态的节点树Tree；
    生成规划路径：反向搜索节点树Tree，获得机器人的规划路径。
  2. 根据权利要求1所述的非规则形状移动机器人的路径规划方法，其特征在于，所述构建工作空间网络的步骤具体为：
    将机器人的二维工作空间均匀划分成拓扑的工作空间网络T，工作空间网络T的每个节点包括自由空间和障碍物占据两个状态，自由空间的节点函数值设定为0，障碍物占据的节点函数值设定为1，工作空间网络T中的节点T i会与周边8个节点相邻，将工作空间网络T进行生物神经动力学迭代计算，构成工作空间的生物神经网络S，生物神经网络S上节点的函数值描述了该节点所在位置的环境信息，生物神经网络S则描述了整个工作空间的环境信息。
  3. 根据权利要求2所述的非规则形状移动机器人的路径规划方法，其特征在于，所述生物神经网络S迭代计算公式为：
    Figure PCTCN2020086533-appb-100001
    公式一中，x i为第i个神经元节点S i的在t时刻的函数值，A、B、D为设定的非负恒定参数，且有A>B>D，I i为生物神经网络节点S i的外界激励，且有
    Figure PCTCN2020086533-appb-100002
    障碍物占据节点的外界激励为I i＝E，自由空间节点的外界激励为I i＝-E，且有E>>B，I j为第j个相邻的神经元，j＝1，2…8，式一中的阈值函数定义为：[a] +＝max{a,0}、[a] -＝max{-a,0}，相邻神经元的连接权重w ij＝1/d，d为第i和第j个神经元节点之间的距离，k＝8为周边8个相邻的神经元。
  4. 根据权利要求1所述的非规则形状移动机器人的路径规划方法，其特征在于，所述构建机器人形状核矩阵步骤具体为：
    将移动机器人所占据区域扩展成规则矩形或正方形区域，并进行均匀划分，构成核矩阵 W，矩阵的每个元素包括自由空间和机器人占据两个状态，自由空间的元素值设定为0，机器人占据的元素值设定为1，将机器人在工作空间内，间隔45°进行旋转，构建反映机器人8个方向的核矩阵W θ，θ＝1，2，…，8。
  5. 根据权利要求4所述的非规则形状移动机器人的路径规划方法，其特征在于，所述构建机器人的移动状态矩阵和转动状态矩阵的步骤具体为：
    将8组反映机器人形状的核矩阵W θ分别与生物神经网络S进行卷积计算，得到8个一层卷积网络H 1,θ，H 1,θ网络中每个节点的输出h 1,θ(x，y)表示机器人是否能够以θ方向放置于工作空间节点(x，y)，且h 1,θ(x，y)＝0表示能够放置，h 1,θ(x，y)＝-1表示不能够放置；
    构建8个3×3的核矩阵M θ，用M θ分别对一层卷积网络H 1,θ进行二次卷积计算，得到8个二层卷积网络H 2,θ，H 2,θ网络中每个节点的输出h 2,θ(x，y)表示机器人是否能够以θ方向从当前节点(x，y)行使进入相邻的节点，且h 2,θ(x，y)＝0表示能够行驶进入，h 2,θ(x，y)＝-1表示不能够行驶进入；
    提取h 2,θ(x，y)，θ＝1，2，…，8，构成一个1×8的移动状态矩阵TSM(x，y)，描述了机器人在当前节点的移动输出状态，即是否能够从节点(x，y)行使进入相邻的8个节点的状态；
    提取h 1，θ(x，y)，θ＝1，2，…，8，构成一个1×8的转动状态矩阵RSM(x，y)，它由‘0’或‘-1’组成，描述机器人在节点(x，y)的转动状态。
  6. 根据权利要求5所述的非规则形状移动机器人的路径规划方法，其特征在于，所述第一层卷积计算方法为：
    Figure PCTCN2020086533-appb-100003
    公式二中，s为核矩阵W θ所在生物神经网络S中所占据的子空间集合，‘*’代表卷积计算，b表示安全系数，采用符号激活函数f(·)，定义为
    Figure PCTCN2020086533-appb-100004
  7. 根据权利要求1所述的非规则形状移动机器人的路径规划方法，其特征在于，所述构建节点树的步骤具体为：
    以机器人的初始状态，搜索机器人能够进入的相邻神经元信息，构成一组节点信息集合P step，搜索步长step＝step+1，判断目标信息是否属于信息集合P step，如果目标信息属于信息集合P step，则停止搜索，如果目标信息不属于信息集合P step，则更新已经搜索的节点树Tree，并将节点信息集合P step设定为当前状态，进行下一步搜索，直到搜索到目标信息为止；
    由每一步长的节点集合P step，step＝1，2，…p，构成一个从初始状态到目标状态的节点树Tree。
  8. 根据权利要求1所述的非规则形状移动机器人的路径规划方法，其特征在于，所述生成规划路径具体步骤为：
    根据所述节点树Tree，输入机器人的初始位置(x 0，y 0)和方向θ 0，输入目标位置(x p，y p)和方向θ p；
    从目标信息节点出发，搜索与之相连的父节点，构成父节点集合，确定路径节点，更新路径节点Path，然后进行循环搜索，直到步长step＝1，由每一步长step＝1，2，…p搜索到的节点构成连接目标位置和初始位置的路径节点Path，即规划路径。
  9. 根据权利要求8所述的非规则形状移动机器人的路径规划方法，其特征在于，所述确定路径节点的确定具体为：如果父节点集合中有水平或垂直相连的父节点，则将其作为机器人行走路径节点，如果没有，则寻找与机器人当前方向转动角度较小的父节点，并将其作为机器人行走路径节点。
  10. 根据权利要求1所述的非规则形状移动机器人的路径规划方法，其特征在于，所述生物神经网络S的节点所描述的实际空间与反映机器人形状的核矩阵W θ的节点所描述的实际空间相等，机器人放置于生物神经网络S中所占据的是一个与核矩阵W大小相等的子网络，将反映机器人形状的核矩阵W θ放置于生物神经网络S中，能够保证W θ与S的节点位置一一对应。
PCT/CN2020/086533 2020-03-09 2020-04-24 一种非规则形状移动机器人的路径规划方法 Ceased WO2021179409A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
GB2105342.6A GB2593058B (en) 2020-03-09 2020-04-24 Path planning method for irregularly shaped mobile robot

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010157313.3 2020-03-09
CN202010157313.3A CN111474925B (zh) 2020-03-09 2020-03-09 一种非规则形状移动机器人的路径规划方法

Publications (1)

Publication Number Publication Date
WO2021179409A1 true WO2021179409A1 (zh) 2021-09-16

Family

ID=71747225

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/086533 Ceased WO2021179409A1 (zh) 2020-03-09 2020-04-24 一种非规则形状移动机器人的路径规划方法

Country Status (3)

Country Link
CN (1) CN111474925B (zh)
GB (1) GB2593058B (zh)
WO (1) WO2021179409A1 (zh)

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114898013A (zh) * 2022-07-15 2022-08-12 深圳市城市交通规划设计研究中心股份有限公司 一种交通等时圈生成方法、电子设备及存储介质
CN115291597A (zh) * 2022-05-30 2022-11-04 上海仙途智能科技有限公司 基于双向混合a*算法的路径规划方法及装置、终端
CN115903794A (zh) * 2022-11-03 2023-04-04 上海思岚科技有限公司 一种多边形机器人的路径生成方法及设备
CN117804463A (zh) * 2024-03-01 2024-04-02 山东力聚机器人科技股份有限公司 一种智慧勘查机器人和路径规划方法
CN119159587A (zh) * 2024-11-06 2024-12-20 南京林业大学 用于轨道客车侧墙板的全覆盖腻子喷涂路径规划方法
CN120008614A (zh) * 2025-04-16 2025-05-16 盛视科技股份有限公司 一种窄道脱困路径规划方法及智能机器人

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113156933B (zh) * 2020-12-30 2022-05-03 徐宁 一种机器人行进控制系统和方法
CN115115078B (zh) * 2021-03-23 2025-03-21 中移(苏州)软件技术有限公司 一种路径搜索方法及装置、设备
CN115016491B (zh) * 2022-06-24 2025-12-16 江苏大学 一种基于rrt的非规则形状移动机器人的路径规划方法
CN116922398B (zh) * 2023-09-15 2023-12-22 华侨大学 一种绳索机器人及其路径规划方法和装置

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110004581A1 (en) * 2009-07-02 2011-01-06 Palo Alto Research Center Incorporated Depth-First Search For Target Value Problems
CN104050390A (zh) * 2014-06-30 2014-09-17 西南交通大学 一种基于可变维粒子群膜算法的移动机器人路径规划方法
CN106843211A (zh) * 2017-02-07 2017-06-13 东华大学 一种基于改进遗传算法的移动机器人路径规划方法
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
CN110057362A (zh) * 2019-04-26 2019-07-26 安徽理工大学 有限单元地图的移动机器人路径规划方法
CN110426044A (zh) * 2019-08-09 2019-11-08 华南理工大学 一种基于凸集计算和优化蚁群算法的避障路径规划方法

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5112666B2 (ja) * 2006-09-11 2013-01-09 株式会社日立製作所 移動装置
CN100491084C (zh) * 2007-07-03 2009-05-27 北京控制工程研究所 一种基于二元环境信息的移动机器人局部路径规划方法
CN101441736B (zh) * 2007-11-21 2010-09-15 新乡市起重机厂有限公司 汽车起重机机器人路径规划方法
CN104317292A (zh) * 2014-09-16 2015-01-28 哈尔滨恒誉名翔科技有限公司 一种复杂形状机器人避碰路径的方法
US11017679B2 (en) * 2017-01-13 2021-05-25 Skydio, Inc. Unmanned aerial vehicle visual point cloud navigation
US20190184561A1 (en) * 2017-12-15 2019-06-20 The Regents Of The University Of California Machine Learning based Fixed-Time Optimal Path Generation
CN109814557A (zh) * 2019-01-23 2019-05-28 西北工业大学 一种全局规划器主导的机器人路径规划方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110004581A1 (en) * 2009-07-02 2011-01-06 Palo Alto Research Center Incorporated Depth-First Search For Target Value Problems
CN104050390A (zh) * 2014-06-30 2014-09-17 西南交通大学 一种基于可变维粒子群膜算法的移动机器人路径规划方法
CN106843211A (zh) * 2017-02-07 2017-06-13 东华大学 一种基于改进遗传算法的移动机器人路径规划方法
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
CN110057362A (zh) * 2019-04-26 2019-07-26 安徽理工大学 有限单元地图的移动机器人路径规划方法
CN110426044A (zh) * 2019-08-09 2019-11-08 华南理工大学 一种基于凸集计算和优化蚁群算法的避障路径规划方法

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115291597A (zh) * 2022-05-30 2022-11-04 上海仙途智能科技有限公司 基于双向混合a*算法的路径规划方法及装置、终端
CN114898013A (zh) * 2022-07-15 2022-08-12 深圳市城市交通规划设计研究中心股份有限公司 一种交通等时圈生成方法、电子设备及存储介质
CN115903794A (zh) * 2022-11-03 2023-04-04 上海思岚科技有限公司 一种多边形机器人的路径生成方法及设备
CN117804463A (zh) * 2024-03-01 2024-04-02 山东力聚机器人科技股份有限公司 一种智慧勘查机器人和路径规划方法
CN119159587A (zh) * 2024-11-06 2024-12-20 南京林业大学 用于轨道客车侧墙板的全覆盖腻子喷涂路径规划方法
CN120008614A (zh) * 2025-04-16 2025-05-16 盛视科技股份有限公司 一种窄道脱困路径规划方法及智能机器人
CN120008614B (zh) * 2025-04-16 2025-07-08 盛视科技股份有限公司 一种窄道脱困路径规划方法及智能机器人

Also Published As

Publication number Publication date
CN111474925A (zh) 2020-07-31
GB2593058A (en) 2021-09-15
GB202105342D0 (en) 2021-05-26
CN111474925B (zh) 2021-09-10
GB2593058B (en) 2022-09-07

Similar Documents

Publication Publication Date Title
CN111474925B (zh) 一种非规则形状移动机器人的路径规划方法
Yaoming et al. A newly bio-inspired path planning algorithm for autonomous obstacle avoidance of UAV
Dao et al. A multi-objective optimal mobile robot path planning based on whale optimization algorithm
CN111880561B (zh) 城市环境下基于改进鲸鱼算法的无人机三维路径规划方法
WO2018176596A1 (zh) 基于权重改进粒子群算法的无人自行车路径规划方法
CN115033016A (zh) 一种异构无人集群编队避障方法及系统
CN110375761A (zh) 基于增强蚁群优化算法的无人驾驶车辆路径规划方法
CN115729238A (zh) 一种用于移动机器人自主避障的动态路径规划方法
CN116774733B (zh) 一种多无人机覆盖路径规划方法
CN108985516B (zh) 基于元胞自动机的室内路径规划方法
CN111930121A (zh) 一种室内移动机器人的混合路径规划方法
CN111552298A (zh) 一种基于鼠脑海马空间细胞的仿生定位方法
Lei et al. A nature-inspired algorithm to adaptively safe navigation of a Covid-19 disinfection robot
Liu et al. Improved gray wolf optimization algorithm integrating A* algorithm for path planning of mobile charging robots
CN114063614A (zh) 基于机载单目视觉感知的多小型机器人系统混合队形快速收敛方法
CN111531542B (zh) 一种基于改进a*算法的机械臂无碰撞路径规划方法
Zhao et al. Improved biological neural network approach for path planning of differential drive agricultural robots with arbitrary shape
Shi et al. Path planning for mobile robots in complex environments based on improved ant colony algorithm.
Yang et al. Sampling-efficient path planning and improved actor-critic-based obstacle avoidance for autonomous robots
Huadong et al. A path planning method of robot arm obstacle avoidance based on dynamic recursive ant colony algorithm
Mou et al. Three-dimensional area coverage with UAV swarm based on deep reinforcement learning
Wu et al. Based on improved bio-inspired model for path planning by multi-AUV
CN114237226A (zh) 一种群机器人复杂环境区域覆盖下的导航脱困方法
CN118625850A (zh) 基于未知环境下四旋翼飞行器时间最优避障在线规划方法
CN119087814A (zh) 一种面向动目标的协同定位状态寻优决策方法及装置

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 202105342

Country of ref document: GB

Kind code of ref document: A

Free format text: PCT FILING DATE = 20200424

WWE Wipo information: entry into national phase

Ref document number: 2105342.6

Country of ref document: GB

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

Ref document number: 20924022

Country of ref document: EP

Kind code of ref document: A1

WWG Wipo information: grant in national office

Ref document number: 2105342.6

Country of ref document: GB

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20924022

Country of ref document: EP

Kind code of ref document: A1

122 Ep: pct application non-entry in european phase

Ref document number: 20924022

Country of ref document: EP

Kind code of ref document: A1