A path planning method for a metamorphic wheeled mobile robot

Abstract

The application relates to a path planning method for a deformable wheeled mobile robot. The method is based on an ESDF map, abstracts the deformable wheeled mobile robot into a geometric model, searches an initial path of the robot through a hybrid A* algorithm, performs branch collision of the hybrid A* algorithm through incremental collision detection, then converts the initial path into a B-spline curve containing position and deformation degree information of the robot over time through fitting, and then calculates motion feasibility between control points of the optimized B-spline curve, motion smoothness and distance information of the geometric model to obstacles. The application aims to provide a smoother and safer path for the deformable wheeled mobile robot, lightens the search process, and enables the robot to quickly complete a path planning task in an unknown map.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of path planning, in particular to a path planning method for a transformable wheeled mobile robot. BACKGROUND

[0002] With the development of science and technology, robot technology is also rapidly advancing, and there are mature application cases in many fields. Unlike the past human-computer interaction mode or simple automation, the current robot application is developing towards complete automation to improve work efficiency and reduce work risk. The path planning of the robot plays an important role in automation, and whether it is a household robot or an industrial robot, it cannot do without path planning. People's interest in transformable mobile robots is gradually deepening, but the use of such robots is usually still in the prospecting stage, or a large amount of resources are needed to achieve the expected goal. How to use the transformable mobile robot is a problem to be solved at present. SUMMARY

[0003] Therefore, the present application aims to provide a path planning method for a transformable wheeled mobile robot, which is lightweight in the search process and can enable the robot to quickly complete the path planning task in an unknown map, thereby making up for the inadaptability of traditional path search algorithms for transformable mobile robots.

[0004] To achieve the above-mentioned purpose, the present application adopts the following technical solution: a path planning method for a transformable wheeled mobile robot, specifically comprising the following steps:

[0005] Step S1: converting an indoor three-dimensional environment into an ESDF map, and abstracting a physical model of the transformable wheeled mobile robot into a geometric model, which is dynamically adjusted according to the transformation state of the robot; combining the degree of transformation of the transformable mobile robot, using an incremental collision detection method to perform branch collision detection, and then performing path search by using an improved hybrid A* algorithm;

[0006] Step S2: discretely sampling the path points of the improved hybrid A* algorithm and fitting them into a B-spline curve, which contains the position and transformation degree information of the robot over time;

[0007] Step S3: taking the obtained B-spline curve as the optimization object, and optimizing the path in terms of path smoothness, motion speed feasibility, and safety distance between the geometric model of the transformable robot and the obstacle.

[0008] In a preferred embodiment, the robot of the improved hybrid A* algorithm in step S1 is expressed in the form of a map model:

[0009] The indoor three-dimensional environment is converted into a three-dimensional ESDF grid map, wherein, in addition to storing information on whether an obstacle is occupied in each grid, distance information of the grid from the nearest obstacle is also stored in a non-occupied grid.

[0010] The real physical model of the transformable wheeled mobile robot is abstracted into a geometric model, which can truly reflect the position information, direction information and deformation degree information of the transformable wheeled mobile robot, and can always reflect the real-time state of the real transformable mobile robot, so that the center of the obtained geometric model of the transformable mobile robot is projected on the ground of the map as a base point to improve the hybrid A* algorithm search.

[0011] In a preferred embodiment, the branch search mode of the improved hybrid A* algorithm in step S1 is:

[0012] Based on the kinematics of the robot, a fixed time step is used as the search resolution to search for the path, and at the same time, the acceleration, angular acceleration and deformation acceleration restrictions are additionally added to each fixed time step in the hybrid A* algorithm to ensure the smoothness of the path and the feasibility of the motion, and the heuristic function of each branch is judged as:

[0013] f(n) = g(n) + h(n)

[0014] g(n) = |Δh| + λ1*Δt + g(n-1)

[0015] h(n) = λ1*t est

[0016] Wherein, f(n) represents the cost estimation function from the initial state of the transformable wheeled mobile robot to the target state through the change of state n;

[0017] g(n) represents the actual cost of the transformable mobile robot from the initial state to state n; |Δh| is the absolute value of the deformation degree difference between the current state n and the previous state n-1, h is the deformation degree of the robot, Δt is the time difference between the current state n and the previous state n-1, and g(n-1) represents the cumulative cost of the robot from the initial state to the previous state n-1;

[0018] h(n) is the best estimated cost from state n to the target state, and the time t of the robot to the end point is predicted through the Dubins curve est The adjustment coefficient λ1 is multiplied to adjust the influence of time on the cost estimation function, the larger the value of λ1, the better the time optimality of the path, and vice versa, the smaller the value of λ1, the better the deformation smoothness of the path.

[0019] In a preferred embodiment, the collision detection of the branch in the improved hybrid A* algorithm in step S1 is:

[0020] The branch collision detection process is to detect each state in the search branch one by one, specifically: in each state of the deformed mobile robot, the newly added volume difference of the geometric model of the current state robot compared with the geometric model of the previous state is calculated, and whether the newly added volume difference collides with the obstacle is detected at the same time.

[0021] In a preferred embodiment, the B-spline fitting method in step S2 is:

[0022] The initial path calculated by the improved hybrid A* algorithm in step S1 is discretely sampled to obtain path discrete points, and then the initial state, terminal state, speed and acceleration limit of the geometric model of the deformed mobile robot are used as conditions to perform B-spline curve fitting on the discrete points to obtain a k-th B-spline curve. The expression of the curve function B(t) is as follows:

[0023]

[0024] Where P i is the control point, and the information of each control point contains {x, y, z, h}, where x, y, and z are the coordinates of the robot, and N i,k+1 (t) is the basis function of the B-spline curve;

[0025] The first derivative of the curve function with respect to time gives the speed information of the deformed wheeled mobile robot, and the second derivative gives the acceleration information of the deformed mobile robot.

[0026] In a preferred embodiment, the optimization method in step S3 is:

[0027] According to the control points of the B-spline curve fitted in step S2, there are N+1 control points, and the first and last k control points are fixed, and the N+1-2k intermediate control points are adjusted.

[0028] For the deformed mobile robot, the overall penalty function f total is:

[0029] f total = λ2f s + λ3f c + λ4f v

[0030] λ2λ3λ4 are the adjustment coefficients of each penalty function.

[0031] f s is the path smoothness penalty function, which functions to make the path smoother by constraining the change amplitude between the control points of the path, satisfying the smoothness requirement of the robot motion, and its equation is:

[0032]

[0033] P i+1 Indicates the next control point.

[0034] f c Let be the penalty function for the robot's distance from obstacles, used to improve path safety; its equation is:

[0035]

[0036] D(P i This reflects the robot's position at control point P. i Distance information to obstacles is obtained through control point P. i The resulting geometric model in step S1 is filled with M points selected within it. The relative positions of these filling points remain unchanged within the geometric model, thus comprehensively reflecting the spatial distribution characteristics of the geometric model. The coordinates of the filling points are taken as p. j Let i = 1, 2, ..., M, and the nearest distance of each point to the obstacle be d(p). j If the robot is at control point P, then the robot is at control point P. i Information about the nearest obstacle can be normalized to

[0037]

[0038] Therefore, this penalty function takes into account both the robot's position adjustment and deformation degree adjustment to reduce the risk of collision;

[0039] f v This is a speed feasibility penalty function used to constrain the speed of path control points. When the speed of a control point exceeds a preset speed limit, f... v The function value will increase significantly, thus guiding path optimization towards meeting the speed limit. Its equation is:

[0040]

[0041] v is the robot's moving speed, v h V is the robot's deformation speed. iμ For the robot at control point P i The speed of movement or the speed of deformation.

[0042] Therefore, the goal of path optimization for this deformable mobile robot is to adjust the values ​​of the control points {x,y,z,h} along the path to make the overall penalty function f total To reach the minimum value; calculate P for each control point. i For the penalty function f totalThe gradient direction of the path is used as the basis for optimization adjustment, thereby generating an optimized path that meets the requirements of path smoothness, obstacle safety, and speed feasibility.

[0043] The values of the coefficients λ2, λ3, and λ4 are adjusted according to actual requirements. If the path needs to be smoother, the proportion of the λ2 coefficient needs to be increased. If the path needs to be safer, the proportion of the λ3 coefficient needs to be increased. If the path needs to comply with the kinematic laws of the robot, the proportion of the λ4 coefficient needs to be adjusted.

[0044] Compared with the prior art, the present application has the following beneficial effects:

[0045] (1) The present application abstracts the physical model of the transformable wheeled mobile robot into a geometric model to improve the accuracy and reliability of the robot's expression in the map. Through this modeling method, the position, direction, and shape characteristics of the robot in different transformable states can be more accurately reflected, thereby optimizing the accuracy and effectiveness of path planning;

[0046] (2) The incremental collision detection method is used, and the degree of transformation of the robot is considered at each path calculation, which can improve the accuracy and real-time performance of collision detection. This method can more accurately evaluate the relative position of the robot in different transformable states with obstacles, ensuring that the robot does not collide with obstacles during transformation, and improving the safety of the robot.

[0047] (3) When optimizing the path, the path smoothness, speed feasibility, and safe distance of the robot from obstacles are comprehensively optimized, so that the path is not only smoother, but also ensures that the robot's movement speed is within an acceptable range, and the path does not collide with obstacles. This comprehensive consideration allows the robot to perform tasks in more complex environments and effectively avoids conflicts and problems that may arise from single-target optimization. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 is a planning schematic diagram of the preferred embodiment of the present application;

[0049] Figure 2 is a schematic diagram of the physical model of the transformable mobile robot being abstracted into a geometric model in the preferred embodiment of the present application;

[0050] Figure 3 is a collision detection schematic diagram of the preferred embodiment of the present application;

[0051] Figure 4 is a path optimization schematic diagram of the preferred embodiment of the present application, wherein, Figure 4 (a) is a position adjustment path optimization schematic diagram, Figure 4 (b) is a transformable degree adjustment path optimization schematic diagram;

[0052] Figure 5 is a search and optimization schematic diagram of a preferred embodiment of the present application, wherein (a) is in a state of not starting the search, wherein the green grid represents the obstacle grid observed by the camera in real time, and the red 8 dots represent the geometric model of the robot; (b) is in a stage of just starting the search, wherein the red path is the optimized B-spline curve, that is, the local path; (c) is after a period of time, the robot executes the subsequent local path; (d) is the case of passing through a narrow road section. DETAILED DESCRIPTION

[0053] The present application will be further described below in conjunction with the drawings and embodiments.

[0054] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present application. Unless otherwise indicated, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application pertains.

[0055] It should be noted that the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the exemplary embodiments according to the present application; as used herein, the singular form is intended to include the plural form unless the context clearly indicates otherwise, and furthermore, it should be understood that when the terms "comprise" and / or "include" are used in the specification, there is a feature, step, operation, device, component and / or combination thereof.

[0056] As shown in Figures 1-5 , the path planning of a deformable wheeled mobile robot according to the present embodiment comprises the following steps: S1, converting an indoor three-dimensional environment into an ESDF map, and abstracting a physical model of the deformable wheeled mobile robot into a geometric model, which can be dynamically adjusted according to the deformation state of the robot; combining the degree of deformation of the deformable mobile robot, and using an incremental collision detection method to perform branch collision detection, thereby improving a hybrid A* algorithm to perform path search.

[0057] S2, discretely sampling the path points of the improved hybrid A* algorithm and fitting them into a B-spline curve, which contains the position and deformation degree information of the robot over time.

[0058] S3, taking the obtained B-spline curve as the optimization object, and optimizing the path in terms of path smoothness, motion speed feasibility, and safety distance between the geometric model of the deformable robot and the obstacle.

[0059] S1 specifically comprises:

[0060] The robot in the map model expression manner of the improved hybrid A* algorithm:

[0061] The indoor three-dimensional environment is converted into a three-dimensional ESDF grid map, wherein, in addition to storing information whether an obstacle is occupied in each grid, distance information of the grid from the nearest obstacle is also stored in a non-occupied grid;

[0062] A real physical model of a transformable wheeled mobile robot is abstracted into a geometric model, as shown in the figure, the geometric model can truly reflect the position information, direction information and deformation degree information of the transformable wheeled mobile robot, and can always reflect the real-time state of the real transformable mobile robot, so that the center of the geometric model of the transformable mobile robot is projected on the ground of the map as a base point to improve the hybrid A* algorithm search; Figure 2

[0063] The branch search mode of the improved hybrid A* algorithm is as follows:

[0064] Based on the kinematics of the robot, a fixed time step is used as the search resolution to search for the path, and at the same time, the acceleration, angular acceleration and deformation acceleration restrictions are additionally added to each fixed time step in the algorithm to ensure the smoothness of the path and the feasibility of the motion, and the heuristic function of each branch is judged as follows:

[0065] f(n) = g(n) + h(n)

[0066] g(n) = |Δh| + λ1*Δt + g(n-1)

[0067] h(n) = λ1*t est

[0068] Wherein, f(n) represents the cost estimation function of the transformable wheeled mobile robot from the initial state to the target state through the change of state n.

[0069] g(n) represents the actual cost of the transformable mobile robot from the initial state to state n; |Δh| is the absolute value of the difference in deformation degree between the current state n and the previous state n-1, h is the deformation degree of the robot, Δt is the time difference between the current state n and the previous state n-1, and g(n-1) represents the cumulative cost of the robot from the initial state to the previous state n-1.

[0070] h(n) is the best estimated cost of state n to the target state, and the time t of the robot to the end point is predicted through the Dubins curve est multiplied by the adjustment coefficient λ1, the adjustment coefficient λ1 can adjust the influence of time on the cost estimation function, the larger the value of λ1, the better the time optimality of the path, and vice versa, the smaller the value of λ1, the better the deformation smoothness of the path;

[0071] The collision detection of the branch in the improved hybrid A* algorithm is as follows: ​

[0072] The branch collision detection process is to detect each state in each search branch one by one, as shown in the following figure: Figure 3 Specifically, in each state of the deformed mobile robot, the volume difference newly added by the geometric model of the current state robot compared with the geometric model of the previous state is calculated, and whether the newly added volume difference collides with the obstacle is detected at the same time. In the corresponding figure, state 2 needs to detect only the red grid in addition to state 1.

[0073] Step S2 specifically includes:

[0074] Discretely sampling the initial path calculated by the improved hybrid A* algorithm of step S1 to obtain path discrete points, and then fitting a k-th B-spline curve to the discrete points under the conditions of the initial state, the terminal state, the speed and acceleration limit of the geometric model of the deformed mobile robot, to obtain a k-th B-spline curve. The expression of the curve function is as follows:

[0075]

[0076] Wherein, is the control point, the information of each control point contains {x, y, z, h}, wherein x, y, z are the coordinates of the robot, and h is the deformation degree of the robot, N i,k+1 (t) is the basis function of the B-spline curve;

[0077] The first order derivative of the curve function with respect to time can obtain the speed information of the deformed wheeled mobile robot, and the second order derivative can obtain the acceleration information of the deformed mobile robot;

[0078] Step S3 specifically includes:

[0079] According to the control points of the B-spline curve fitted in step S2, there are N+1 control points, and the first and last k control points are kept unchanged, and the N+1-2k control points in the middle are adjusted;

[0080] For the deformed mobile robot, the overall penalty function f total of the path optimization is:

[0081] f total = λ2f s + λ3f c + λ4f v

[0082] λ2, λ3, λ4 are the adjustment coefficients of each penalty function.

[0083] f s is the path smoothness penalty function, which functions to make the path smoother by constraining the change amplitude between the control points of the path, so as to meet the smoothness requirement of the robot motion, and its equation is:

[0084]

[0085] P i+1 Indicates the next control point.

[0086] f c Let be the penalty function for the robot's distance from obstacles, used to improve path safety; its equation is:

[0087]

[0088] D(P i This reflects the robot's position at control point P. i Distance information to obstacles is obtained through control point P. i The resulting geometric model in step S1 is filled with M points selected within it. The relative positions of these filling points remain unchanged within the geometric model, thus comprehensively reflecting the spatial distribution characteristics of the geometric model. The coordinates of the filling points are taken as p. j Let i = 1, 2, ..., M, and the nearest distance of each point to the obstacle be d(p). j If the robot is at control point P, then the robot is at control point P. i Information about the nearest obstacle can be normalized to

[0089]

[0090] Therefore, this penalty function considers both the robot's position adjustment and deformation degree adjustment to reduce collision risk, such as Figure 4 As shown, where, Figure 4 (a) is a schematic diagram of the location adjustment path optimization. Figure 4 (b) is a schematic diagram of the path optimization for adjusting the degree of deformation.

[0091] f v This is a speed feasibility penalty function used to constrain the speed of path control points. When the speed of a control point exceeds a preset speed limit, f... v The function value will increase significantly, thus guiding path optimization towards meeting the speed limit. Its equation is:

[0092]

[0093] v is the robot's moving speed, v h V is the robot's deformation speed. iμ For the robot at control point P i The speed of movement or the speed of deformation.

[0094] Therefore, the goal of path optimization for this deformable mobile robot is to adjust the values ​​of the control points {x,y,z,h} along the path to make the overall penalty function ftotal reaches the minimum value. Calculate each control point P i the gradient direction of the penalty function f total , as the basis for optimization adjustment, so as to generate a method of optimizing the path that meets the requirements of path smoothness, obstacle safety and speed feasibility, Figure 5 for the overall process, Figure 5 (a) is in the state of not starting the search, where the green grid represents the obstacle grid observed by the camera in real time, and the red 8-point represents the geometric model of the robot; Figure 5 (b) is in the stage of just starting the search, where the red path is the optimized B-spline curve, that is, the local path; Figure 5 (c) is after a period of time, the robot executes the subsequent local path; Figure 5 (d) is through the narrow road section, which can be clearly seen from the difference with Figure 5 (a) deformation degree.

[0095] According to the actual requirements, the values of the coefficients λ2, λ3 and λ4 can be reasonably adjusted. If the path needs to be smoother, the proportion of the λ2 coefficient needs to be increased; if the safety performance of the path needs to be enhanced, the proportion of the λ3 coefficient needs to be increased; and if the path needs to be more in line with the kinematics of the robot, the proportion of the λ4 coefficient needs to be adjusted.

[0096] In summary, according to steps S1, S2 and S3, the path of the transformable mobile robot can be optimized based on the improved hybrid A* algorithm, the B-spline curve, and the three purposes of path smoothness, motion speed feasibility and safe distance between the geometric model of the transformable robot and the obstacle, thereby obtaining a path planning method applicable to the transformable wheeled mobile robot.

Claims

1. A method for path planning of a metamorphic wheeled mobile robot, characterized in that, Specifically comprising the following steps: Step S1: converting the indoor three-dimensional environment into an ESDF map, and abstracting the physical model of the transformable wheeled mobile robot into a geometric model that dynamically adjusts with the transformable state of the robot; By combining the degree of deformation of the transformable mobile robot, an incremental collision detection method is used for branch collision detection, and thus an improved hybrid A* algorithm is used for path search; Step S2: discretely sampling the path points of the improved hybrid A* algorithm and fitting them into a B-spline curve, which contains the position and deformation degree information of the robot over time; Step S3: taking the obtained B-spline curve as the optimization object, and optimizing the path in terms of path smoothness, motion speed feasibility, and safety distance between the geometric model of the transformable robot and the obstacle; The optimization method of step S3: According to the control points of the B-spline curve obtained in step S2, a total of control points are obtained, and control points at the beginning and the end are fixed, and control points in the middle are adjusted. For the morphing mobile robot, the overall penalty function for path optimization is: F = F1 + F2 + F3 + F4 are adjustment coefficients for each penalty function; is the path smoothness penalty function, which is used to make the path smoother by constraining the variation between the control points of the path, and its equation is: represents the next control point; A penalty function for the robot distance from the obstacle, to improve the safety of the path, whose equation is: reflecting the robot at the control point distance information of the distance to the obstacle, through the control point the geometric model in step S1, selecting filling points inside the geometric model, the relative positions of these filling points in the geometric model remain unchanged, and the filling points can fully reflect the distribution characteristics of the geometric model in space, taking the filling point coordinates as the nearest distance of each point to the obstacle is , the robot at the control point distance information of the nearest obstacle is normalized to Therefore, the penalty function simultaneously considers the position adjustment and deformation degree adjustment of the robot to reduce the risk of collision; is a speed feasibility penalty function used to constrain the speed of the path control points, when the speed of a certain control point exceeds the pre-set upper speed limit, the function value will increase significantly, thus guiding the path optimization to adjust its equation in the direction that meets the speed limit. It is a movement speed of the robot, a deformation speed of the robot, a movement speed or a deformation speed of the robot at the control point ; Therefore, the goal of path optimization for the morphing mobile robot is to minimize the total penalty function by adjusting the values of the control points of the path; the gradient direction of the penalty function at each control point is calculated and used as the basis for optimization adjustment, thereby generating an optimized path that meets the requirements of path smoothness, obstacle safety, and speed feasibility. The value of the coefficient is adjusted according to the actual requirements If the path is to be made smoother, the proportion of the coefficient needs to be increased; If the path is to be made safer, the coefficient ratio needs to be increased If the path is to be made more in line with the kinematics of the robot, the coefficient ratio needs to be adjusted .

2. The path planning method for a metamorphic wheeled mobile robot according to claim 1, wherein The robot in the improved hybrid A* algorithm in step S1: Convert the indoor three-dimensional environment into a three-dimensional ESDF grid map, where each grid stores information about whether it is occupied by an obstacle, and in the non-occupied grid, it also stores the distance information from the nearest obstacle; The real physical model of the transformable wheeled mobile robot is abstracted into a geometric model that can truly reflect the position information, direction information and deformation degree information of the transformable wheeled mobile robot, and always reflects the real-time state of the real transformable mobile robot, and the center of the obtained geometric model of the transformable mobile robot is projected on the ground of the map as the base point for improved hybrid A* algorithm search.

3. The path planning method for a metamorphic wheeled mobile robot according to claim 1, wherein, The branch search method of the improved hybrid A* algorithm in step S1: Based on the kinematics of the robot, a fixed time step is used as the search resolution to search for the path, and in the hybrid A* algorithm, an acceleration, angular acceleration, and deformation acceleration constraint is added to each fixed time step to ensure the smoothness and feasibility of the motion, and the heuristic function of each branch is determined as: wherein, represents a cost estimation function from a deformed wheeled mobile robot by a change of state from an initial state to a target state; represents the actual cost of the morphing mobile robot from the initial state to the state ; is the absolute value of the difference in the degree of morphing between the current state and the previous state , is the degree of morphing of the robot, is the time difference between the current state and the previous state , represents the cumulative cost of the robot from the initial state to the previous state ; is the state is the best estimate of the cost to the goal state, which predicts the time of the robot to the end point by Dubins curve is multiplied by the adjustment coefficient is the adjustment coefficient is the effect of the adjustment time on the cost estimation function, The larger the value is, the better the path time optimality is, and vice versa, The smaller the value is, the better the path deformation smoothness is.

4. The path planning method for a metamorphic wheeled mobile robot according to claim 1, wherein, The collision detection of the branch in the improved hybrid A* algorithm in step S1: The branch collision detection process is to detect each state in the search branch one by one, specifically: in each state of the transformable mobile robot, calculate the newly added volume difference of the geometric model of the current state robot compared to the geometric model of the previous state, and simultaneously detect whether the newly added volume difference collides with the obstacle.

5. The path planning method for a metamorphic wheeled mobile robot according to claim 1, wherein, The B-spline fitting method in step S2: The initial path calculated by the improved mixed A* algorithm of step S1 is discretely sampled to obtain path discrete points, and a B-spline curve fitting is performed on the discrete points under the conditions of the initial state, the terminal state, the speed and the acceleration limit of the deformed mobile robot geometric model to obtain a secondary B-spline curve The expression of the secondary B-spline curve function is as follows: wherein for the control points, the information of each control point contains wherein is the coordinate of the robot, is the basis function of the B-spline curve; The first derivative of the curve function with respect to time gives the speed information of the transformable wheeled mobile robot, and the second derivative gives the acceleration information of the transformable mobile robot.

Images