A robot autonomous exploration method based on bio-inspired optimization

Abstract

The application belongs to the field of robot autonomous exploration, and discloses a robot autonomous exploration method based on biological heuristic optimization. Based on the grey wolf optimization algorithm, the method combines the robot autonomous exploration strategy driven by the boundary, and aims to propose a random exploration strategy determined by the algorithm of the robot path point. The method is divided into local path point selection for obstacle avoidance and global path point selection for guiding exploration. Firstly, the light value of the laser radar which is not reflected from the obstacle is expressed as a boundary point and stored in the stack; the boundary points are divided into three groups according to the distance cost, the boundary point with the average distance cost in each group is found, and the point found by the grey wolf optimization algorithm is taken as the global path point to guide the robot exploration; when the robot goes to the global path point, five local points are used to change the direction of the robot to the local path point, so that the robot can avoid obstacles; the local path point and the global path point are switched to guide the robot autonomous exploration.

Description

Technical Field

[0001] This invention relates to the field of autonomous robot exploration, and specifically describes a method for autonomous robot exploration based on bio-inspired optimization. Background Technology

[0002] The exploration strategy based on robot autonomous exploration in unknown environments aims to enable robots to cover the entire unknown environment as completely as possible. It studies how to use data acquired by the robot's own onboard sensors to autonomously explore the targets and scenes to be explored in unknown environments, that is, to mainly solve the problems of "where to go, how to go, and how to cover as much of the environment as possible".

[0003] In recent years, most strategies for autonomous exploration by robots have been based on two-dimensional environments. When applied to complex and unknown three-dimensional environments, their efficiency and performance are not ideal.

[0004] Classic autonomous exploration strategies fall into two main categories: boundary-based exploration strategies and NBV-based exploration strategies.

[0005] The exploration strategy based on the NBV (next-best-view) concept involves iteratively finding the observation point location with the richest perceptual information, and then moving the robot to these observation points to complete the exploration task. However, during the exploration process, the robot may pay less attention to areas with less information gain and easily overlook small areas.

[0006] Traditional boundary-based exploration strategies define the boundary as the area where the known and unexplored regions meet, and guide the robot to move towards the boundary to continuously increase the area covered by the environment. This type of strategy has the following problems: as the area of ​​the explored region increases, the efficiency of boundary detection decreases, causing the robot to fail to detect boundary changes in a timely manner, resulting in repeated exploration. Greedy strategies that head towards the nearest boundary are prone to getting trapped in local optima. Furthermore, if such deterministic algorithms experience collisions, the failure will recur at the same time and place until the environment, the robot, or the algorithm changes. This necessitates that robots try more flexible heuristic techniques. Summary of the Invention

[0007] This invention provides a bio-inspired optimization-based autonomous exploration method for robots. By combining boundary-based thinking and the gray wolf optimization algorithm, it abandons the traditional deterministic exploration method based on boundary thinking, enabling robots to conduct more flexible autonomous exploration.

[0008] The technical solution adopted by the present invention to achieve the above objectives is as follows:

[0009] A bio-inspired optimization-based method for autonomous robot exploration includes the following steps:

[0010] 1) When the robot encounters an obstacle while moving forward, it selects a local path point and performs obstacle avoidance based on the local path point;

[0011] 2) Calculate and select global path points to guide the robot forward. When the robot encounters an obstacle, select a local path point.

[0012] Step 1) includes the following steps:

[0013] 1.1) When the robot encounters the first obstacle while moving straight, determine whether the length of the emitted laser beam is equal to the threshold V. dist The size of the path is considered; if it is less than the value of the path, a local path point is calculated; otherwise, the global path point is used.

[0014] 1.2) Check for obstacles around local waypoints and select and proceed to reachable local waypoints.

[0015] The local path points are five target points used by the robot to change its direction from the current direction of travel of the global path points. Their coordinates on the x-axis and y-axis are defined in a coordinate system with the robot as the origin. They are respectively:

[0016]

[0017]

[0018] The robot is surrounded by five local waypoints, all located L meters in front of it. yes The angles of the five local path points relative to the x-axis are π / 4, π / 2, π, -π / 2, and -π / 4, respectively. Let be the current robot pose in the j-th iteration.

[0019] Step 2) includes the following steps:

[0020] 2.1) Calculate the cost of all boundary points and sort the boundary points in descending order according to the cost to obtain a set of boundary point sequences;

[0021] 2.2) An improved gray wolf optimization algorithm is used to find global path points;

[0022] 2.3) The robot moves to the global path point and determines whether the distance to the global path point is less than the maximum coverage area of ​​the sensors on the robot. If it is less, return to step 2.1); otherwise, continue to the global path point.

[0023] 2.4) When the Grey Wolf optimization algorithm reaches the maximum number of iterations, the search for a global path point ends.

[0024] The cost of the boundary points for:

[0025]

[0026] in, These are the x-axis and y-axis values ​​of the k-th boundary point, respectively, where the boundary point is the boundary point of the robot's LiDAR coverage area.

[0027] Step 2.2) specifically refers to:

[0028] The first three groups of m boundary points in the boundary point sequence set are set as the α, β, and δ wolf packs in the Grey Wolf Optimization Algorithm, and the cost of the boundary points is... As the distance cost, the average boundary point in each group is found based on the distance cost. The initial positions of the α, β, and δ wolf packs are used respectively, and the output of the gray wolf optimization algorithm is used as the global path point.

[0029] The improved gray wolf optimization algorithm is based on the original gray wolf optimization algorithm, but replaces the original convergence factor with a nonlinear convergence factor. The nonlinear convergence factor is:

[0030]

[0031] Among them, a s and a e Here, q represents the initial and final values ​​of the parameter, q is the nonlinear adjustment parameter, and k is the number of iterations of the Grey Wolf optimization algorithm. max This represents the maximum number of iterations.

[0032] The present invention has the following beneficial effects and advantages:

[0033] Compared to existing autonomous robot exploration technologies, this invention includes a strategy for switching between global and local pathpoints, as well as an autonomous exploration method that selects global pathpoints using a heuristic algorithm. The proposed method is more flexible and adaptable to various environments. This invention estimates global pathpoints based on boundary points and the robot's current position. During exploration, it switches to local pathpoints for obstacle avoidance due to the presence of obstacles. This switching between global and local pathpoints defines the exploration strategy of this invention. The introduction of a bio-inspired optimization algorithm, namely the Grey Wolf Optimization Algorithm, to select global pathpoints further enhances the flexibility of autonomous exploration. Attached Figure Description

[0034] Figure 1 A flowchart of the present invention.

[0035] Figure 2a A diagram of the local path point selection module of this invention.

[0036] Figure 2bA schematic diagram of obstacle detection at local path points according to the present invention.

[0037] Figure 2c A schematic diagram of local path point obstacle avoidance in this invention. Detailed Implementation

[0038] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0039] like Figure 1 The diagram shows a flowchart of a bio-inspired optimization-based autonomous robot exploration method. This embodiment uses a four-wheel drive ground robot, which is a prior art technology. The bio-inspired optimization-based autonomous robot exploration method includes the following steps:

[0040] Step 1: The robot travels straight and encounters the first obstacle. It then determines the length of the laser beam relative to the threshold V. dist The size; if it is less than, go to step 2; if it is not less than, go to step 10.

[0041] Step 2, calculate local path points;

[0042] Step 3: Check for obstacles around the local waypoints;

[0043] Step 4: Select and proceed to an accessible local waypoint;

[0044] Step 5: The robot reaches a local path point; the local path point module ends and proceeds to step 6.

[0045] Step 6: Calculate the cost of all boundary points;

[0046] Step 7: Sort the boundary points in descending order based on the cost;

[0047] Step 8: Use the Grey Wolf optimization algorithm to find global path points;

[0048] Step 9: The robot proceeds to the global waypoint;

[0049] Step 10: Determine if the distance to the global path point is less than the sensor's maximum coverage area; if less, proceed to step 6; otherwise, proceed to step 11.

[0050] Step 11, continue to the global waypoint;

[0051] Step 12: The algorithm terminates when it reaches its maximum number of iterations.

[0052] In one embodiment, the calculation of local path points in step 2 is as follows:

[0053] Figure 2 shows the local path point selection module. Obstacles exist in all environments, and the robot should pass safely. However, the sensor detection range has a limited field of view for planning obstacle avoidance ahead. It requires some safe scenarios to avoid obstacles within the detection range. Therefore, as shown in Figure 2(a), five local points v are used to change the robot's current direction from the global path point to the local path point. Equations (1) and (2) calculate the selection of local path points on the x-axis and y-axis, respectively:

[0054]

[0055]

[0056] The robot is surrounded by n local points (n=5), located L meters in front of the robot. yes Relative to the angles between the x-axis, the corresponding angles for v5 are π / 4, v4 is π / 2, v3 is -π / 2, and v2 is -π / 4. The current robot pose in the j-th iteration. Robot's orientation. The robot can rotate 360 ​​degrees depending on the direction of travel.

[0057] In one embodiment, step 3 involves checking obstacles around the local waypoint as follows:

[0058] Figure 2(a) illustrates the representation of local pathpoint locations at a certain distance and angle around the robot. Since a local pathpoint is a pixel in size, it appears as a single point in eight directions. Additional checks were performed to ensure that the V shown in Figure 2(b) was within acceptable limits. dist Detect obstacles within a range of / 2.

[0059] In one embodiment, step 4 involves selecting and traversing a reachable local waypoint as follows:

[0060] Figure 2(c) shows that points cannot be selected in this situation. Because their position and reflected light provide information about obstacles, the next local point will be... The robot selects from among several safe local paths. Once a safe local pathpoint is finally chosen, the robot needs some time to avoid obstacles before continuing towards the global pathpoint.

[0061] The global waypoint selection module allows the robot to travel straight without any designated global waypoints at the start of the exploration process. The initial direction of travel is unimportant because the entire space is unknown. Once the robot reaches an obstacle, the first global waypoint is generated.

[0062] In one embodiment, the cost of calculating all boundary points in step 6 is as follows:

[0063] The calculation of global pathpoints depends on boundary points. Each boundary point has its own location. and cost Each iteration updates and stores the data in the stack.

[0064] cost It is estimated based on the current robot position using Euclidean metrics.

[0065]

[0066] in These are the x-axis and y-axis values ​​of the k-th boundary point, respectively.

[0067] Additionally, boundary points that meet the following conditions should be removed:

[0068] (1) Boundary points on obstacles.

[0069] (2) Boundary points that have already been explored.

[0070] (3) The distance to the robot is less than the range of the sensor.

[0071] In one embodiment, step 8 uses the Grey Wolf optimization algorithm to find the global path point as follows:

[0072] Boundary points The values ​​are modified according to the robot's drive. The global pathpoints, based on the input boundary point array, assign the first three groups of m boundary points in the boundary point sequence as α, β, and δ wolf packs, and find the average boundary point in each group based on distance cost. The improved gray wolf optimization algorithm was used to find it.

[0073] The average boundary point for each group is obtained by dividing the sum of the costs of all points in each group by the number of points.

[0074] The improved gray wolf optimization algorithm adopts the same gray wolf pack hunting mechanism as the original gray wolf optimization algorithm: wolf pack hierarchy, surrounding prey, and preying on prey.

[0075] The wolf pack hierarchy divides wolf hunting into four levels: α, β, δ, and ω.

[0076] The act of surrounding prey is defined as follows:

[0077]

[0078]

[0079] Equation (4) represents the distance between a gray wolf and its prey, and Equation (5) is the gray wolf's position update formula. These are the position vectors of the prey and the gray wolf, respectively, where k is the current iteration number. The influence coefficient is calculated using the following formulas:

[0080]

[0081]

[0082] in It is the convergence factor. The original convergence factor decreases linearly from 2 to 0 as the number of iterations increases. It is a vector of two one-dimensional components with values ​​in the range [0, 1].

[0083]

[0084] Finally, the wolf pack, led by α, β, and δ wolves, gradually approaches its prey, their positions dynamically changing, until the hunt is successful. This process can be represented by a formula:

[0085]

[0086]

[0087]

[0088] In equation (9), represent the distances between other individuals in the wolf pack and α, β, and δ wolves, respectively; These represent the current positions of α, β, and δ wolves, respectively. This indicates the current position of the individual gray wolf.

[0089] Equation (10) These represent the positions that other individuals in the wolf pack need to adjust due to the influence of α, β, and δ wolves, respectively. This determined its final location.

[0090] The improved gray wolf optimization algorithm specifically includes:

[0091] In the original gray wolf optimization algorithm, the convergence factor significantly impacts both the global search capability and the local exploration capability. Global search capability refers to the ability of the wolf pack to explore other unexplored areas to prevent them from getting trapped in local optima. When |A|>1, the wolf pack needs to search the entire space for prey. Local exploration capability represents the accuracy within a small area. When |A|<1, the wolf pack aims to surround and attack prey; local exploration capability also determines the convergence speed. Therefore, the convergence factor plays a significant role. The convergence factor 'a' decreases linearly from 2 to 0 with increasing iterations. To achieve better performance, a nonlinear convergence factor can be used, which decreases slowly in the early stages of iteration, facilitating global search, and accelerates convergence in later stages to find a better solution more quickly. The nonlinear convergence factor is as follows:

[0092]

[0093] In equation (12), a s and a e These are the initial and final values ​​of the parameter, respectively, and q is the nonlinear adjustment parameter (an integer greater than 1).

[0094] Furthermore, the robot random autonomous exploration method based on gray wolf optimization specifically includes:

[0095] Initialize 'a', and determine the maximum number of iterations.

[0096] Calculate A1, A2, A3, C1, C2, C3;

[0097] When the distance of the ray returned by the sensor is less than the maximum coverage area of ​​the sensor, three sets of points, α, β, and δ, will be generated, with m points in each set;

[0098] Find the average boundary point in each group based on distance cost.

[0099] Calculate using formula (9) Then, calculate using formula (10) to obtain Finally, the global pathpoint is obtained using formula (11). The entire exploration process will be completed under the strategy of switching between local and global pathpoints until the Grey Wolf optimization algorithm iterates to the maximum number of times.

Claims

1. A method for autonomous robot exploration based on bio-inspired optimization, characterized in that, Includes the following steps: 1) When the robot encounters an obstacle while moving forward, it selects a local path point and performs obstacle avoidance based on the local path point; 2) Calculate and select global path points to guide the robot forward. When the robot encounters an obstacle, select a local path point. Step 2) includes the following steps: 2.1) Calculate the cost of all boundary points and sort the boundary points in descending order according to the cost to obtain a set of boundary point sequences; 2.2) An improved gray wolf optimization algorithm is used to find global path points; 2.3) The robot moves to the global path point and determines whether the distance to the global path point is less than the maximum coverage area of ​​the sensors on the robot. If it is less, return to step 2.1); otherwise, continue to the global path point. 2.4) When the Grey Wolf optimization algorithm reaches the maximum number of iterations, the search for a global path point ends; The improved gray wolf optimization algorithm is based on the original gray wolf optimization algorithm, but replaces the original convergence factor with a nonlinear convergence factor. The nonlinear convergence factor is: ； in, a s and a e Here, represents the initial and final values ​​of the parameter, and q is the nonlinear adjustment parameter. To optimize the number of iterations in the gray wolf algorithm, This represents the maximum number of iterations.

2. The method for autonomous robot exploration based on bio-inspired optimization according to claim 1, characterized in that, Step 1) includes the following steps: 1.1) When the robot encounters the first obstacle while moving straight, determine whether the length of the emitted laser beam is equal to the threshold V. dist The size of the path is considered; if it is less than the value of the path, a local path point is calculated; otherwise, the global path point is used. 1.2) Check for obstacles around local waypoints and select and proceed to reachable local waypoints.

3. The method for autonomous robot exploration based on bio-inspired optimization according to claim 2, characterized in that, The local path points are five target points used by the robot to change its direction from the current direction of travel of the global path points. Their coordinates on the x-axis and y-axis are defined in a coordinate system with the robot as the origin. , They are respectively: ； ； The robot is surrounded by five local waypoints, all located L meters in front of it. yes The angles of the five local path points relative to the x-axis are as follows: , , , , , , , Let be the current robot pose in the j-th iteration.

4. The method for autonomous robot exploration based on bio-inspired optimization according to claim 1, characterized in that, The cost of the boundary points for: ； in, These are the x-axis and y-axis values ​​of the k-th boundary point, respectively, where the boundary point is the boundary point of the robot's LiDAR coverage area.

5. The method for autonomous robot exploration based on bio-inspired optimization according to claim 1, characterized in that, Step 2.2) specifically refers to: The first three groups of m boundary points in the boundary point sequence set are set as the α, β, and δ wolf packs in the Grey Wolf Optimization Algorithm, and the cost of the boundary points is... As the distance cost, the average boundary point in each group is found based on the distance cost. , , , respectively, are used as the initial positions of the α, β, and δ wolf packs, and the output of the gray wolf optimization algorithm is used as the global path point.

Images