A method for planning a sequence of footfalls for a biped robot in a truss environment

Abstract

The application discloses a truss environment-oriented two-legged robot crawling landing point sequence planning method and belongs to the technical field of robots; the method first constructs a truss undirected graph topology model based on the spatial connection relationship of truss joints and members, takes nodes to represent the truss joints, takes edges to represent the members and takes the member length as the edge weight; the A star algorithm is adopted on the topology model to search for the node sequence and the member sequence of the robot from the starting member to the target member; subsequently, the axial position of the member is described through a normalized member parameter model, and the non-connected feasible landing area set of the robot foot end is represented in combination with One-Hot coding; on this basis, the least number of steps required for the robot to reach the target position is taken as an optimization target, a mixed integer quadratic constraint programming model is established, the landing point sequence is solved, and the optimal landing point sequence satisfying the truss topology constraint and the robot motion constraint is obtained. The application can realize the landing point sequence planning of a two-legged truss crawling robot on a complex truss structure.

Description

Technical Field

[0001] This invention belongs to the field of robotics technology and relates to a method for planning the sequence of foot landing points during the crawling process of a bipedal robot in a truss environment. Background Technology

[0002] With the continuous development of robotics technology and intelligent planning methods, the autonomous movement and operation capabilities of robots in complex structural environments have become a research hotspot. Among them, bipedal truss crawling robots are a type of robot that moves by alternately grasping truss members with both legs, enabling them to move and manipulate on the surfaces of spatial trusses and structural frames. These robots typically rely on truss structures as supports and motion carriers, achieving crawling and structural traversal through discrete footholds and transitions between members. They are widely used in fields such as spatial structure assembly, inspection and maintenance, and hazardous environment operations.

[0003] In truss structures, the structure typically consists of numerous joints and members, whose connections can be abstracted as a graph structure model. Truss joints can be represented as graph nodes, and member connections as graph edges. For such discrete topological environments, the A*Star algorithm is a typical heuristic graph search algorithm that combines actual path costs with heuristically estimated costs to improve search efficiency while ensuring path reachability.

[0004] On the other hand, robot motion planning and sequential decision-making problems often involve the joint optimization of discrete choice variables and continuous parameter variables. For example, in the landing planning process, it is necessary to simultaneously determine the discrete landing area selection and the continuous spatial position parameters. Mixed Integer Quadratic Constrained Programming (MIQCP) is a class of mathematical optimization methods that can handle both integer variables and quadratic constraints, and can be used to describe optimization problems involving combined decision-making and nonlinear constraints.

[0005] Based on the above technical background, how to combine truss topology modeling, graph search methods and mixed integer optimization methods to realize robot foot planning for truss structure environments has become an important research direction in related technical fields.

[0006] This invention proposes a method for planning the landing point sequence of a bipedal robot crawling in a truss environment. It involves path search in an undirected graph, representation of discrete feasible regions, and mixed integer optimization. This method can plan a suitable landing point sequence for the crawling robot in a complex discrete truss environment, supporting complex task planning and motion planning. Summary of the Invention

[0007] This invention proposes a method for planning the landing point sequence of a bipedal robot crawling in a truss environment. The technical solution is as follows: First, an undirected graph topological representation model of the truss structure is constructed, with nodes describing truss joints and edges describing link connections. Based on this topological model, the A*-star search algorithm is used to obtain the node paths and link paths of the robot from the starting link to the target link. Then, a normalized link parameterization model is used to describe the axial positions of the links, and One-Hot encoding is combined to characterize the set of non-connected feasible landing regions at the robot's feet. Further, a landing point sequence optimization model is established with the minimum number of steps as the optimization objective, and a mixed-integer quadratic constraint programming method is used to solve it, thereby obtaining a landing point sequence that satisfies both truss topological constraints and motion feasibility constraints. The invention process is as follows:

[0008] 1. Representing Truss Topology Using Undirected Graphs: Model the truss topology by taking the truss joint number, spatial coordinates, member number, length, and topological connection relationship as inputs to construct an undirected graph topology model of the truss. Undirected graph nodes represent truss joints, and undirected graph edges represent members connecting joints. Member lengths are used as edge weights to establish a topological representation for path planning.

[0009] 2. The A*Star algorithm is used to plan the node sequence and link sequence for the robot to crawl from the link at the starting position to the link at the target position, and the direction vector of each link is generated from the node sequence: The node sequence and link sequence of the robot's crawling process are planned using the A*Star algorithm. First, the links at the starting position of the robot's two feet and the link at the target position are determined. The A*Star algorithm is used to search for paths on the undirected graph topology model to obtain the node sequence connecting the starting link and the target link and the corresponding link sequence. Second, the missing path endpoints are filled in with nodes to form a continuous node sequence, and the direction vector of the link is defined based on the node sequence.

[0010] 3. Considering the infeasible footing region near the truss joint, the feasible footing region is defined using One-Hot encoding and a normalized member parameter model: The set of non-connected feasible regions is defined. First, a normalized parameter model is established based on the starting spatial coordinates and length of each member in the member sequence to describe the axial position of the member. Then, considering the infeasible region in the neighborhood of the truss joint caused by the robot's geometric constraints, the feasible footing intervals on each member are calculated, and One-Hot encoding is used to represent the set of non-connected feasible regions at the robot's foot.

[0011] 4. Using the minimum number of steps to reach the target position as the optimization objective, Mixed Integer Quadratic Constrained Programming (MIQCP) is used to solve for the landing point sequence: With the minimum number of steps required for the robot to reach the target position as the optimization objective, an MIQCP model is established by combining the feasible landing area constraint and the motion continuity constraint, and the optimal landing point sequence that satisfies the constraints is obtained by solving the model. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the bipedal truss crawling robot of the present invention.

[0013] Figure 2 This is a flowchart illustrating the landing point planning process of the present invention;

[0014] Figure 3 This is a schematic diagram of the truss undirected graph of the present invention;

[0015] Figure 4 The node sequence and member sequence planning results of this invention;

[0016] Figure 5 This is a schematic diagram of the set of non-connected feasible regions in this invention;

[0017] Figure 6 This is a schematic diagram of the optimal landing point sequence of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the following examples provide a more detailed description of the invention. It should be noted that the specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention.

[0019] The Figure 1 This diagram illustrates the crawling motion of a bipedal truss-crawling robot on a truss. The robot needs to determine the landing point for each step on the truss to reach the target position through the alternating movement of its two feet. The specific implementation process consists of four steps, as follows: Figure 2 As shown:

[0020] The first step is to model the truss topology.

[0021] Using the truss joint number, spatial coordinates, member number, length, and topological connection relationship as input, define an undirected graph of the truss. ,in Represents the nodes of the graph. The edges of the graph are represented by the weights of the bars. Sure;

[0022] Taking a "straight" truss consisting of 52 rods and 28 truss joints as an example, the undirected graph topology modeling results of the truss are shown below. Figure 3 It consists of 24 rods with a length of 832mm and 28 rods with a length of 432mm.

[0023] The second step is to plan the node sequence and truss sequence during the crawling process.

[0024] Determine the starting positions of the robot's two feet. Member number Target positions at both feet Member number .

[0025] The A*Star algorithm is used to search for node and link sequences on the undirected graph topology model, with the goal of minimizing the sum of edge weights, to obtain the node sequence connecting the starting link and the target link. and the corresponding rod sequence .

[0026] The missing path endpoints are filled in with nodes to form a continuous node sequence, and the direction vector of the member is defined based on the node sequence.

[0027] Taking the truss established in step one as an example, first determine the starting positions of the robot's two feet. The number of the rod in question and target location The number of the rod in question The A*-star algorithm is used to search for node and member sequences on the undirected truss graph established in step one. After supplementing the nodes of the starting and ending members, the node sequence is obtained as follows: and the rod sequence is ,like Figure 4 As shown.

[0028] The third step is to define the set of non-connected feasible regions.

[0029] according to Each rod The starting spatial coordinates , direction vector and normalization parameters Construct a normalized parameter model for the member, see formula. ,in , . formula It can describe any position along the axial direction of the rod.

[0030]

[0031] Considering the infeasible region in the neighborhood of the truss joints due to robot geometric constraints, calculate the feasible footing intervals on each member. .

[0032] The set of non-connected feasible regions at the robot's foot is represented by One-Hot encoding. Each pole is labeled using a formula. The robot's feet are restricted to a single set of feasible regions, in which... This refers to the number of rods.

[0033]

[0034] The members described in step two are normalized and a feasible region set is defined. A 40mm radius around the joint of each member is defined as an infeasible foothold area. The feasible foothold areas for the two lengths of members are calculated as follows: , Encoded via One-Hot Each pole is labeled to obtain A diagram illustrating non-connected feasible regions is shown below. Figure 5 As shown, the black dashed lines represent feasible areas, and the thick black solid lines represent infeasible areas.

[0035] The fourth step is to plan the optimal landing point sequence.

[0036] The robot's motion process is modeled using the MIQCP model. First, the maximum number of steps required for the robot to move to the target position is set. The robot moves to the target position by alternating bipedal movements according to a predetermined sequence of links. Due to physical constraints, the robot can only move one foot in one step.

[0037] Based on the normalized link parameter model and One-Hot encoding identification method described in step three, the robot's first... Left foot position From the formula Obtain, among which and The third step described above The starting position and direction vector of the root rod. and For the first The first step of the left foot One-Hot encoding and normalization parameters of the root rod. Based on the One-Hot encoding constraints and upper and lower limits of a single feasible interval described in step three, the robot's foot can only be located within one feasible interval at a time, restricting the foot position of each step to only within a non-connected feasible landing area.

[0038]

[0039] Similarly, the robot's first Step right foot position From the formula get.

[0040]

[0041] and Must meet To ensure that the robot always moves forward.

[0042] The robot's movement must meet physical constraints, including that the distance between the robot's two feet must always be within an allowable range, not exceeding the robot's link length limit, nor being too small to cause the robot's feet to collide. This is determined by the formula... describe.

[0043]

[0044] In adjacent steps, the robot's step length cannot exceed the link length limit, as specified by the formula. Description. Because the robot can move forward by twice the radius of a circle centered on one foot, with the other foot also moving forward. This method allows for a larger step size, with an upper limit of the step size. .

[0045]

[0046] Add terminal constraints to the model, constraining the robot to move from the starting position at each of its two feet to the target position at each foot, as shown by the formula. describe.

[0047]

[0048] In each step, the robot moves only one foot, while the other foot remains stationary. Without loss of generality, assuming the robot moves its left leg first, the robot's position update method can be obtained as follows: .

[0049]

[0050] The optimization objective of the model is defined as the robot reaching the target position in the fewest steps, as shown in the formula. As shown. Among them. For the first The One-Hot code value of the last link at the left foot end (the link where the target position is located). For the first Normalized parameters of the last link at the left foot end (the link where the target position is located). The meaning is: when the first When the left foot is not on the target member, its value is 0; when the first step... When the left foot reaches the target member, its value is The same applies to the right end.

[0051]

[0052] Using the data described in steps two and three as input to the optimization model, set... , , Solving using existing mathematical programming solvers, the optimal landing point sequence is as follows: Figure 6 As shown in Table 1, hollow dotted lines represent the sequence of foot placements on the left end, and solid dotted lines represent the sequence of foot placements on the right end. The robot takes a total of 17 steps to reach the target position, and the foot placements for each foot are shown in Table 1.

[0053] Table 1. Optimal landing point sequence for the robot

[0054] 0 (0,4600,0) (0,4500,0) 1 (0,4200,0) (0,4500,0) 2 (0,4200,0) (0,3880,0) 3 (0,3560,0) (0,3880,0) 4 (0,3560,0) (0,3240,0) 5 (0,2920,0) (0,3240,0) 6 (0,2920,0) (0,2600,0) 7 (0,2280,0) (0,2600,0) 8 (0,2280,0) (0,1960,0) 9 (121.59,1664,0) (0,1960,0) 10 (121.59,1664,0) (432,1586,0) 11 (432,1266.25,0) (432,1586,0) 12 (432,1266.25,0) (432,946.25,0) 13 (432,832,298.91) (432,946.25,0) 14 (432,832,298.91) (432,540.99,432) 15 (432,300,432) (432,540.99,432) 16 (432,300,432) (432,200,432) 17 (432,100,432) (432,200,432)

[0055] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements can be made without departing from the principle of the present invention, and these improvements should also be considered within the scope of protection of the present invention.

Claims

1. A method for planning the sequence of foot placement points for bipedal robot crawling in a truss environment, characterized in that, The method is as follows: 1) Represent the truss topology using an undirected graph; 2) The A* star algorithm is used to plan the node sequence and link sequence of the robot from the link at the starting position to the link at the target position, and the direction vector of each link is generated through the node sequence; 3) Considering the infeasible foothold area near the truss joint, the feasible foothold area is defined using One-Hot coding and normalized member parameter model; 4) With the goal of reaching the target position in the fewest steps, use mixed integer quadratic constraint programming to solve the landing point sequence.

2. The method for planning the sequence of footholds for a bipedal robot crawling in a truss environment according to claim 1, characterized in that, In step 1), the truss joint number, spatial coordinates, member number, length, and topological connection relationship are used as inputs to construct an undirected graph topological model of the truss. Undirected graph nodes represent truss joints, and undirected graph edges represent members connecting joints. Member lengths are used as edge weights to establish a topological representation for path planning. The undirected graph of the truss is defined. ,in Represents the nodes of the graph. The edges of the graph are represented by the weights of the bars. Sure.

3. The method for planning the sequence of foot placement points for a bipedal robot crawling in a truss environment according to claim 2, characterized in that, Truss joints are mapped to nodes in an undirected graph, containing the three-dimensional spatial coordinates of the joints; members connecting two truss joints are mapped to undirected edges between corresponding nodes, used to describe the adjacency relationship in the truss structure; the weight of each undirected edge is determined by the geometric length of the corresponding member; forming a topological model for member path sequence search and landing point planning.

4. The method for planning the sequence of footholds for a bipedal robot crawling in a truss environment according to claim 1, characterized in that, Step 2) specifically refers to: First, determine the starting link corresponding to the robot's initial position and the target link corresponding to the target position; then determine the initial positions of the robot's two feet. Member number Target positions at both feet Member number ; On the undirected graph topology model of the truss, the link path is searched based on the A* search algorithm to obtain the optimal link sequence and its associated node sequence connecting the starting position and the target position; according to the end relationship between the starting link and the target link, the missing nodes at the beginning and end of the path are filled in; the direction vector of each link is calculated based on the three-dimensional spatial coordinates of the adjacent nodes in the node path to limit the feasible movement direction of the robot's foot.

5. The method for planning the sequence of foot placement points for a bipedal robot crawling in a truss environment according to claim 1, characterized in that, In step 3), the feasible landing area is defined as follows: an axial parametric model of each member is established using the three-dimensional coordinates of its starting point and its length. And by normalizing parameters Describe any position on the axis of the member; considering the infeasible foothold region in the neighborhood of the truss joint caused by the robot's geometric constraints, determine the upper and lower limits of the parameters of the feasible foothold region of each member based on the normalized member parameter model. Discretize and encode the feasible footing regions that are not interconnected on each link, and use One-Hot encoding to represent the feasible region where the robot's foot is currently located.

6. The method for planning the sequence of footholds for a bipedal robot crawling in a truss environment according to claim 1, characterized in that, Step 4) specifically involves: optimizing the robot's journey to the target location by minimizing the number of steps required, constructing an optimization model for landing point sequence planning, with the cost function being... ,in The function is related to the number of steps the robot takes to reach the target position using both feet; the foot placement area selection is represented as an integer decision variable, and the robot's single-step step length and geometric dimensions are represented as quadratic constraints to establish a mixed integer quadratic constraint programming model; the robot can only move one foot at a time, and when the left foot moves, the right foot remains in the same position. The position update can be represented as: as well as The same principle applies to position updates when the right foot moves. The optimal sequence of foot placement points satisfying the constraints is obtained by solving the planning model.

Images