Path planning and gait generation methods for lunar robots in unstructured environments

Abstract

This invention discloses a method for lunar robot path planning and gait generation in unstructured environments, comprising: constructing a triangular mesh terrain reconstruction model that preserves the original terrain undulation features; planning the optimal global movement path for the robot with minimal undulation based on the triangular mesh terrain reconstruction model; determining the optimal landing point based on the optimal global movement path with minimal undulation, and generating a quadrupedal crawling gait. This invention solves the problems of omnidirectionality, continuity, coordination, and terrain adaptability of robot crawling in unstructured environments, improves the robot's autonomous decision-making ability during movement, and enables the robot to reach the target location efficiently and safely.

Description

Technical Field

[0001] This invention belongs to the field of space robot kinematics technology, and in particular relates to a method for path planning and gait generation of lunar robots in unstructured environments. Background Technology

[0002] After landing on the lunar surface with the Chang'e-7 probe, the multi-legged robot will conduct exploration missions in the complex polar terrain. This requires stable movement and turning capabilities in uneven terrain to achieve functions such as walking, climbing, obstacle crossing, and turning. To enable the robot to autonomously crawl and transfer in the unstructured environment of the lunar polar region, the primary challenges are planning the robot's path points and footholds, and switching gait modes, to ensure the robot's autonomy, adaptability, and stability. Traditional multi-legged robot gait generation methods still have some shortcomings, such as gait self-locking, gait discontinuity, and lack of intelligence. Summary of the Invention

[0003] The technical problem solved by this invention is to overcome the shortcomings of the prior art and provide a method for lunar robot path planning and gait generation in unstructured environments. This method aims to solve the problems of omnidirectionality, continuity, coordination, and terrain adaptability of robot crawling in unstructured environments, improve the robot's autonomous decision-making ability during movement, and enable the robot to reach the target location efficiently and safely.

[0004] To address the aforementioned technical problems, this invention discloses a method for lunar robot path planning and gait generation in unstructured environments, comprising:

[0005] Construct a triangular mesh terrain reconstruction model that preserves the original terrain undulation characteristics;

[0006] Based on the triangular mesh terrain reconstruction model, the optimal global movement path of the robot with the least undulation is planned.

[0007] Based on the optimal global movement path of the robot with minimal undulation, the best footing point is determined and a quadruped crawling gait is generated.

[0008] In the aforementioned method for lunar robot path planning and gait generation in unstructured environments, a triangular mesh terrain reconstruction model that preserves the original terrain undulation features is constructed, including:

[0009] Terrain point cloud data is acquired through 3D laser scanning. A cusp removal-based terrain point cloud data filtering method is used to identify and remove cusp noise points with curvature greater than a preset curvature threshold, resulting in terrain data points P with curvature not exceeding the preset curvature threshold. i,j Where i and j represent the serial numbers of the terrain data points;

[0010] Terrain data point Pi,j Marking these as feature points yields the terrain curvature evaluation function k(P). i,j ):

[0011]

[0012] Where n1 to n8 represent the features P i,j The normal vectors of eight adjacent planes;

[0013] The known terrain is segmented using adjacent equilateral triangles, and feature point P is selected. i,j The equation of the fitted triangle plane is denoted as: (The interpolation fitting is performed.)

[0014] Ax + By + Cz + D = 0

[0015] Where A, B, C, and D represent the coefficients of the plane equation of the triangle, based on the side length d of the equilateral triangle and the z-coordinate of the vertex z. t Sure;

[0016] With d and z t To optimize the variables, the minimum sum of distances from all feature points to the plane containing the equilateral triangle is used as the optimization index J, thus constructing a triangular network terrain reconstruction model that preserves the original terrain undulation characteristics.

[0017] In the above-mentioned lunar robot path planning and gait generation method for unstructured environments, the optimization index J is expressed as follows:

[0018]

[0019] Among them, (x p ,y p ,z p ) represents the three-dimensional coordinates of the feature point.

[0020] In the aforementioned method for lunar robot path planning and gait generation in unstructured environments, based on a triangular mesh terrain reconstruction model, the optimal global movement path for the robot with minimal undulation is planned, including:

[0021] Based on the triangular mesh terrain reconstruction model, the robot's "state-action" decision for each step is evaluated and optimized.

[0022] Based on the evaluation and optimization results, select the "state-action" sequence with the highest cumulative reward;

[0023] Based on the selected "state-action" sequence with the highest cumulative reward, the optimal global movement path of the robot with the least fluctuation is planned.

[0024] In the aforementioned method for lunar robot path planning and gait generation in unstructured environments, a triangular mesh terrain reconstruction model is used to evaluate and optimize the robot's "state-action" decision for each step, including:

[0025] In the triangular mesh terrain reconstruction model, the center coordinates of the triangular cell are defined as state s, representing the robot's position; the action that can be executed in each state is defined as a, corresponding to the robot's movement direction, and the movement step size is determined by the side length d of the triangular cell.

[0026] Based on the Q-value function, the robot's "state-action" decision for each step is evaluated and optimized:

[0027] Q k+1 (s,a)=Q k (s,a)+ω(r+λmaxQ k (s',a')-Q k (s,a))

[0028] Where s' represents the robot's next state, a' represents the robot's next action, k represents the number of iterations, ω represents the learning rate, r represents the learning reward related to the degree of terrain undulation and the proximity to the target, and λ represents the discount factor of the reward over time.

[0029] In the above method for lunar robot path planning and gait generation in unstructured environments, the functional expression of r is as follows:

[0030]

[0031] Where, n Δ1 and n Δ2 P represents the direction vector of adjacent triangular cells in the triangular mesh terrain reconstruction model, used to measure the degree of terrain undulation; start and P now These represent the robot's starting position and current position, respectively; P goal ω1 and ω2 represent the target point location and are used to measure the proximity to the target; ω1 and ω2 are non-negative weighting coefficients that sum to 1.

[0032] In the above-mentioned lunar robot path planning and gait generation method for unstructured environments, the robot is a quadruped robot. The quadruped robot adopts a highly stable crawling gait, with each leg rising and falling in sequence. One leg is called the swing leg, and the other three legs are in a stable support state, called the support leg.

[0033] In the aforementioned method for lunar robot path planning and gait generation in unstructured environments, the quadruped robot follows the optimal global movement path with minimal undulations, adaptively adjusting the swing leg swing sequence according to different turning angles; adaptively adjusting the swing leg stride according to mechanism constraints and uneven ground conditions; and simultaneously, the supporting leg adaptively adjusts the robot's pose and center of gravity trajectory based on the terrain undulations and route changes, using the roll potential energy and stability margin, thereby enabling the quadruped robot to achieve omnidirectional, stable, and coordinated adaptive walking in unstructured environments.

[0034] In the above-mentioned lunar robot path planning and gait generation method for unstructured environments, the optimal landing point is determined based on the robot's optimal global movement path with the least undulation, and a quadruped crawling gait is generated. This includes: the quadruped robot walks in a crawling gait, following a cyclic stepping sequence with optimal stability margin, with each leg switching back and forth between a three-legged support phase and a single-leg swinging phase, and adaptively adjusting the leg swinging sequence according to the direction of the robot's optimal global movement path with the least undulation.

[0035] This invention also discloses a lunar robot path planning and gait generation system for unstructured environments, comprising:

[0036] The model construction module is used to construct a triangular mesh terrain reconstruction model that preserves the original terrain undulation features;

[0037] The path planning module is used to plan the optimal global movement path of the robot with the least undulation based on the triangular mesh terrain reconstruction model.

[0038] The gait generation module is used to determine the optimal foot placement point based on the robot's optimal global movement path with minimal undulation, and generate a quadrupedal crawling gait.

[0039] The present invention has the following advantages:

[0040] (1) This invention discloses a lunar robot path planning and gait generation method for unstructured environments. It reconstructs the three-dimensional unstructured environment based on triangular mesh and uses a reinforcement learning-based intelligent planning method based on the meshed environment to improve the robot's decision-making ability and terrain adaptability in unstructured environments.

[0041] (2) This invention discloses a lunar robot path planning and gait generation method for unstructured environments. The gait is adaptively generated according to the terrain undulation and route changes, which ensures that the robot can walk in an all-round, stable and coordinated manner in unstructured environments.

[0042] (3) This invention discloses a lunar robot path planning and gait generation method for unstructured environments. It adopts a hierarchical motion planning strategy that combines global planning and local planning, which effectively improves motion planning efficiency and ensures that the robot can crawl and transfer efficiently in unstructured environments. It has the advantages of high efficiency, strong passability, good obstacle crossing safety and high motion autonomy.

[0043] (4) This invention discloses a path planning and gait generation method for lunar robots in unstructured environments. By intelligent path planning and adaptive gait optimization, the coordination relationship between multi-legged robots in spatial and temporal order is determined, which ensures the terrain adaptability, turning and obstacle crossing ability of lunar robots in unstructured environments such as complex terrain in polar regions, and can produce positive practical effects.

[0044] (5) This invention discloses a lunar robot path planning and gait generation method for unstructured environments, which can be used to solve the problem of adaptive walking in rugged and obstacle-ridden terrain environments such as other extraterrestrial surfaces and spacecraft surfaces, and provide technical support for the implementation of tasks such as on-site exploration of extraterrestrial planets and transfer and deployment of space surface robots. Attached Figure Description

[0045] Figure 1 This is a flowchart illustrating the steps of a lunar robot path planning and gait generation method for unstructured environments, as described in an embodiment of the present invention.

[0046] Figure 2 This is a schematic diagram of the neighborhood of terrain data points and the normal vectors of adjacent planes in an unstructured environment according to an embodiment of the present invention.

[0047] Figure 3 This is a schematic diagram of the movement path of a lunar robot in an unstructured environment according to an embodiment of the present invention;

[0048] Figure 4 This is a schematic diagram of a lunar robot moving straight and turning in an embodiment of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments disclosed in the present invention will be described in further detail below with reference to the accompanying drawings.

[0050] One of the core ideas of this invention is as follows: a lunar robot path planning and gait generation method for unstructured environments is adopted, which combines global planning and local planning. First, global intelligent path planning is performed based on reinforcement learning to initially ensure the robot's adaptability to unstructured environments. Then, the robot's foot placement is adaptively adjusted according to the current environmental characteristics, and a quadruped crawling gait is generated along the planned path. By adaptively adjusting the leg swing sequence and swing leg stride, the coordination relationship of each leg of the quadruped robot in space and time is determined, thereby effectively improving the efficient mobility, adaptability to unstructured environments, and motion stability of the lunar robot, reducing the spatial dimension of planning, and improving motion planning efficiency.

[0051] like Figure 1 In this embodiment, the lunar robot path planning and gait generation method for unstructured environments includes:

[0052] Step 101: Construct a triangular mesh terrain reconstruction model that preserves the original terrain undulation characteristics.

[0053] In this embodiment, as Figure 2 As shown, the design concept of the triangular mesh terrain reconstruction model is to use conformity preservation as the optimization index, and to segment the unstructured environment using adjacent equilateral triangles to construct a triangular mesh terrain reconstruction model that preserves the original terrain undulation characteristics. Specifically:

[0054] Terrain point cloud data is acquired through 3D laser scanning. Considering that the original data is discrete and irregular, a cusp removal-based terrain point cloud data filtering method is used to identify and remove cusp noise with curvature greater than a preset curvature threshold, resulting in terrain data points P with curvature no greater than the preset curvature threshold. i,j Where i and j represent the serial numbers of the terrain data points.

[0055] Terrain data point P i,j Marking these as feature points yields the terrain curvature evaluation function k(P). i,j ):

[0056]

[0057] Where n1 to n8 represent the features P i,j The normal vectors of 8 mutually adjacent planes.

[0058] The known terrain is segmented using adjacent equilateral triangles, and feature point P is selected. i,j The equation of the fitted triangle plane is denoted as: (The interpolation fitting is performed.)

[0059] Ax + By + Cz + D = 0

[0060] Where A, B, C, and D represent the coefficients of the plane equation of the triangle, based on the side length d of the equilateral triangle and the z-coordinate of the vertex z. t Sure.

[0061] With d and z t To optimize the variables, the minimum sum of distances from all feature points to the plane containing the equilateral triangle is used as the optimization index J, thus constructing a triangular network terrain reconstruction model that preserves the original terrain undulation characteristics.

[0062] The preferred optimization index J is expressed as follows:

[0063]

[0064] Among them, (x p ,y p ,z p ) represents the three-dimensional coordinates of the feature point.

[0065] Step 102: Based on the triangular mesh terrain reconstruction model, the optimal global movement path of the robot with the least undulation is planned.

[0066] In this embodiment, as Figure 3 As shown, in the triangular mesh terrain reconstruction model, based on the robot's current position, near-field environmental perception results, and the specified target point position, the robot's autonomous intelligent path planning algorithm based on reinforcement learning is designed as follows: the terrain undulation degree and target proximity are used as reward values ​​r to feed back to the Q-learning algorithm for Q-value iteration. Furthermore, the robot continuously selects the best direction to move, optimizing the Q-value, and ultimately planning an optimal global movement path for the robot with minimal undulation. Specifically:

[0067] In the triangular mesh terrain reconstruction model, the center coordinates of the triangular cell are defined as state s, representing the robot's position; the action that can be executed in each state is defined as a, corresponding to the robot's movement direction, and the movement step size is determined by the side length d of the triangular cell.

[0068] Based on the Q-value function, the robot's "state-action" decision for each step is evaluated and optimized:

[0069] Q k+1 (s,a)=Q k (s,a)+ω(r+λmaxQ k (s',a')-Q k (s,a))

[0070] Where s' represents the robot's next state, a' represents the robot's next action, k represents the number of iterations, ω represents the learning rate, r represents the learning reward related to the degree of terrain undulation and the proximity to the target, and λ represents the discount factor of the reward over time.

[0071] Based on the evaluation and optimization results, select the "state-action" sequence with the highest cumulative reward.

[0072] Based on the selected "state-action" sequence with the highest cumulative reward, the optimal global movement path of the robot with the least fluctuation is planned, which initially ensures the robot's adaptability to unstructured environments.

[0073] Preferably, the function expression for r is as follows:

[0074]

[0075] Where, n Δ1 and n Δ2 P represents the direction vector of adjacent triangular cells in the triangular mesh terrain reconstruction model, used to measure the degree of terrain undulation; start and P now These represent the robot's starting position and current position, respectively; P goal ω1 and ω2 represent the target point location and are used to measure the proximity to the target; ω1 and ω2 are non-negative weighting coefficients that sum to 1.

[0076] Step 103: Based on the robot's optimal global movement path with minimal undulation, determine the best footing point and generate a quadruped crawling gait.

[0077] In this embodiment, the robot can specifically be a quadruped robot. The quadruped robot employs a highly stable crawling gait, with each leg taking off and landing sequentially. One leg is called the swing leg, while the other three legs are in a stable supporting state, called the support legs. The quadruped robot follows a planned optimal global movement path with minimal undulation, adaptively adjusting the swing leg swing sequence based on different turning angles; adaptively adjusting the swing leg stride based on mechanism constraints and uneven ground conditions; and simultaneously, the support legs adaptively adjust the robot's posture and center of gravity trajectory based on terrain undulations and route changes, using rollover potential energy and stability margin. This enables the quadruped robot to achieve omnidirectional, stable, and coordinated adaptive walking in unstructured environments.

[0078] In this embodiment, as Figure 4 As shown, the quadruped robot walks in a crawling gait. Following the cyclical stepping sequence with optimal stability margin, each leg switches back and forth between a three-legged support phase and a single-leg swinging phase. The leg swinging sequence is adaptively adjusted according to the direction of the robot's optimal global movement path with minimal undulation.

[0079] Let the right front leg of the quadruped robot be 1, the right hind leg be 2, the left hind leg be 3, and the left front leg be 4. The leg swinging sequence is shown in Table 1.

[0080]

[0081] Table 1. Leg Swing Sequence Diagram

[0082] During movement, the robot's center of gravity must remain within the support triangle area with a certain stability margin h. Simultaneously, each leg of the robot must avoid collisions caused by mechanical constraints and uneven ground. To address this, an adaptive adjustment parameter δ is introduced, enabling the robot to adapt to unstructured environments and move with maximum amplitude.

[0083] For the straight-line gait of a quadruped robot, taking the case where leg 3 takes a step while legs 1, 2, and 4 form a support phase to propel the robot's center of gravity, the specific implementation process is as follows:

[0084] The magnitude of the robot's center of gravity shift Δx c It is expressed as follows:

[0085] Δx c =R[δ+(R T (P 2,end -P 2,base ))·[1,0,0] T +h,0,0] T

[0086] Where R represents the robot's body posture; P 2,base and P 2,end δ represents the installation position and foot position of the mobile constraint leg 2, respectively; h represents the stability margin, which is used to ensure that the robot's center of gravity is always in the three-legged support area; δ represents the adaptive adjustment parameter, which enables the robot to avoid collisions that may be caused by mechanism constraints and uneven ground conditions, adapt to unstructured environments, and move with maximum amplitude.

[0087] According to Δx c The position x3 of the landing point of leg 3 is obtained as follows:

[0088] x3(1:2)=(x c (1:2)+Δx c (1:2))×2-x1(1:2)

[0089] x3(3)=f terrain (x3(1:2))

[0090] Where x1 represents the foot position of the supporting leg 1 located diagonally opposite leg 3; the foot height x3(3) of leg 3 is determined according to the triangular mesh terrain model f terrain Sure.

[0091] This will further adjust the robot's center of gravity x after the movement. c Updated to:

[0092]

[0093] In summary, based on the stride maximization index and collision avoidance and obstacle crossing constraints, δ is adaptively adjusted to select the optimal foot placement point, thereby achieving adaptive gait adjustment. At the same time, the fuselage center of gravity is adjusted to change with the height of the foot placement point to adapt to terrain undulations. As for the fuselage attitude, the roll angle is adjusted according to the height difference between the left and right feet, and the pitch angle is adjusted according to the height difference between the front and rear feet, so that the legs have enough swing space to move flexibly.

[0094] For the turning gait of a quadruped robot, the left turn occurs when the hind leg takes its second step, and the right turn occurs when the hind leg takes its third step. Taking the case of a left turn with the hind leg taking its second step as an example, the specific implementation process is as follows: considering that the maximum left turn angle is constrained by the hind leg's first step, the maximum turning angle α of the robot is determined. max To satisfy the constraints:

[0095] ||R z (α max (P) 1,base -x c )+x c -P 1,end ||-L=0

[0096] Among them, R z () represents the rotation matrix about the z-axis; P 1,base P 1,end L and x represent the mounting position, foot position, and leg length of steering restraint leg 1, respectively; c This indicates the position of the machine's center of gravity. Based on the required turning angle α of the planned path, if α < α max Then, leg 2 steps forward while simultaneously propelling the body to rotate by α, followed by legs 1, 4, and 3 stepping forward in sequence; if α > α max Then, during the subsequent third leg step, the body is simultaneously propelled to rotate α-α. max Finally, the robot completed its turning.

[0097] In summary, this invention discloses an autonomous path planning and adaptive gait generation method for lunar robots in unstructured environments. It reconstructs the 3D unstructured environment based on triangulation and utilizes a reinforcement learning-based intelligent planning method to improve the robot's decision-making and terrain adaptability in unstructured environments. The gait is adaptively generated according to terrain undulations and route changes, ensuring the robot's omnidirectional, stable, and coordinated adaptive walking in unstructured environments. A hierarchical motion planning strategy combining global and local planning effectively improves motion planning efficiency, guaranteeing efficient adaptive crawling and transfer in unstructured environments. It features high efficiency, strong obstacle-crossing safety, and high motion autonomy. This invention determines the spatial and temporal coordination relationships of multi-legged robots through intelligent path planning and adaptive gait optimization, ensuring the terrain adaptability, turning, and obstacle-crossing capabilities of lunar robots in complex polar terrain and other unstructured environments, resulting in positive practical effects. The multi-legged robot path planning and gait generation algorithm of this invention can be used to solve the adaptive walking problem in rugged and obstacle-ridden terrain environments such as other extraterrestrial surfaces and spacecraft surfaces, providing technical support for missions such as on-site exploration of extraterrestrial planets and the transfer and deployment of space surface robots.

[0098] Based on the above embodiments, the present invention also discloses a lunar robot path planning and gait generation system for unstructured environments, comprising: a model construction module for constructing a triangular mesh terrain reconstruction model that maintains the original terrain undulation features; a path planning module for planning the optimal global movement path of the robot with the least undulation based on the triangular mesh terrain reconstruction model; and a gait generation module for determining the best footing point based on the optimal global movement path of the robot with the least undulation and generating a quadruped crawling gait.

[0099] As the system implementation corresponds to the method implementation, the description is relatively simple, and relevant details can be found in the description of the method implementation section.

[0100] Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make possible changes and modifications to the technical solutions of the present invention by utilizing the methods and techniques disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solutions of the present invention shall fall within the protection scope of the technical solutions of the present invention.

[0101] The contents not described in detail in this specification are common knowledge to those skilled in the art.

Claims

1. A method for path planning and gait generation of lunar robots in unstructured environments, characterized in that, include: Construct a triangular mesh terrain reconstruction model that preserves the original terrain undulation characteristics; Based on a triangulated terrain reconstruction model, the optimal global movement path for the robot with minimal undulation is planned. This includes: evaluating and optimizing the robot's "state-action" decisions for each step based on the triangulated terrain reconstruction model; selecting the "state-action" sequence with the highest cumulative reward based on the evaluation and optimization results; and planning the optimal global movement path for the robot with minimal undulation based on the selected "state-action" sequence with the highest cumulative reward. The robot is a quadruped, employing a highly stable crawling gait with each leg taking off and landing sequentially. One leg is called the swing leg, while the other three legs are in a stable support state, called the support legs. Along the planned optimal global movement path with minimal undulation, the quadruped adaptively adjusts the swing leg swing sequence according to different turning angles; it also adaptively adjusts the swing leg stride based on mechanism constraints and uneven ground conditions. Simultaneously, the support legs adaptively adjust the robot's pose and center of gravity trajectory based on the terrain undulations and route changes, using flipping potential energy and stability margin, thereby enabling the quadruped robot to achieve omnidirectional, stable, and coordinated adaptive walking in unstructured environments. Based on the optimal global movement path of the robot with the least undulation, the best foot landing point is determined and a quadruped crawling gait is generated; including: the quadruped robot walks in a crawling gait, following the cyclic stepping sequence with the optimal stability margin, with each leg switching back and forth between the three-legged support phase and the single-leg swinging phase, and adaptively adjusting the leg swinging sequence according to the direction of the optimal global movement path of the robot with the least undulation. in: Constructing a triangular mesh terrain reconstruction model that preserves the original terrain undulation features includes: Terrain point cloud data is acquired through 3D laser scanning. A cusp removal-based filtering method is used to identify and remove cusp noise points with curvature exceeding a preset curvature threshold, resulting in terrain data points with curvature not exceeding the preset threshold. ;in, and Indicates the sequence number of the terrain data point; terrain data points Marking these as feature points yields the terrain curvature evaluation function. : in, Representation and feature points The normal vectors of eight adjacent planes; The known terrain is segmented using adjacent equilateral triangles, and feature points are identified. The equation of the fitted triangle plane is denoted as: (The interpolation fitting is performed.) in, , , and The coefficients of the plane equation of a triangle are given by the side length of an equilateral triangle. and the vertex Axis coordinates Sure; by and To optimize the variables, the optimization criterion is to minimize the sum of the distances from all feature points to the plane containing the equilateral triangle. J A triangular mesh terrain reconstruction model that preserves the original terrain undulation characteristics is constructed; optimization indicators are obtained. It is expressed as follows: in, Represents the three-dimensional coordinates of the feature points; Based on the triangulation terrain reconstruction model, the robot's "state-action" decision for each step is evaluated and optimized, including: The center coordinates of the triangular cells in the triangular mesh terrain reconstruction model are defined as the state. , representing the robot's position; the actions that can be performed in each state are defined as follows: Corresponding to the robot's direction of movement, the step size is determined by the side length of the triangular cell. Sure; according to Q The value function evaluates and optimizes the robot's "state-action" decision for each step. in, Indicates the robot's next state. Indicates the robot's next action. k Indicates the number of iterations. Indicates the learning rate. This represents the learning reward related to the degree of terrain undulation and the proximity to the target. This indicates the discount factor that changes the reward over time.

2. The method for lunar robot path planning and gait generation in unstructured environments according to claim 1, characterized in that, The function expression is as follows: in, and This represents the direction vector of adjacent triangular cells in the triangular mesh terrain reconstruction model, used to measure the degree of terrain undulation; and These represent the robot's starting position and current position, respectively. Indicates the location of the target point, used to measure proximity to the target; and These are non-negative weighting coefficients that sum to 1.

3. A lunar robot path planning and gait generation system for unstructured environments, used to implement the method as described in claim 1, characterized in that, include: The model construction module is used to construct a triangular mesh terrain reconstruction model that preserves the original terrain undulation features; The path planning module is used to plan the optimal global movement path of the robot with the least undulation based on the triangular mesh terrain reconstruction model. The gait generation module is used to determine the optimal foot placement point based on the robot's optimal global movement path with minimal undulation, and generate a quadrupedal crawling gait.

Images