Motion planning for biped robots

Abstract

The present invention relates to motion planning for a bipedal robot. According to one embodiment of the present invention, a motion planning method for a bipedal robot includes the steps of: acquiring a discrete map having terrain information corresponding to a target walking area; determining terrain points in the discrete map that the bipedal robot can pass through; and generating an initial landing point sequence based on the terrain points; generating an initial motion trajectory for the bipedal robot based on the discrete map; and optimizing the initial landing point sequence according to the initial motion trajectory and a kinematic model of the bipedal robot based on a predetermined objective function to generate a target landing point sequence. Based on ground height gradient information in the discrete map, it is determined whether a position is passable by the bipedal robot or can be used as a landing point for the bipedal robot, and parameters of the landing points are optimized to obtain a target landing point sequence, so that the bipedal robot can combine terrain characteristics to complete the passing task.

Description

[Technical Field]

[0001] The present invention relates to the field of legged robots, and in particular to motion planning for bipedal robots. [Background technology]

[0002] In recent years, service robots have attracted attention. These robots have certain mobility, sensing, and manipulation capabilities, and can replace traditional human labor in specific fields such as logistics, patrol inspections, and housework, effectively alleviating the labor shortage problem in these fields. Service robots are broadly classified into wheeled and legged robots based on their structure. In the field of legged robots, the problem of generating locomotion gaits for bipedal robots has become a focus of research.

[0003] Currently, gait planning for bipedal robots is achieved by calculating the angles of each joint of the bipedal robot based on a model. That is, a time-varying motion trajectory sequence of the joints is generated, and these trajectory sequences are accurately executed through the joints to complete the movement. For example, a human-like kinematic model and multiple center-of-gravity trajectories are used to switch between gaits at different heights, or a method combining a penguin robot model and a zero moment point (ZMP) trajectory is used to generate a linear walking gait. The underactuated joints between the legs of a bipedal robot and the ground make it difficult to control the motion of these joints, making it difficult to apply gait planning methods based on joint trajectory sequences. Furthermore, gait planning methods for bipedal robots assume that the terrain on which the robot is located is flat. Therefore, gaits generated by these methods cannot be directly applied to terrain with unevenness. Summary of the Invention

[0004] To solve the above problems in the prior art, the present invention provides the following technical solutions.

[0005] According to a first aspect of an embodiment of the present invention, there is provided a motion planning method for a bipedal robot, the motion planning method for a bipedal robot comprising: acquiring a discrete map having topographical information corresponding to a target walking area, determining topographical points on the discrete map that can be passed by the bipedal walking robot, and generating an initial landing point sequence based on the topographical points; generating an initial motion trajectory of the biped walking robot based on the discrete map; and optimizing the initial landing point sequence according to the initial motion trajectory and a kinematic model of the bipedal robot based on a predetermined objective function to generate a target landing point sequence so that the bipedal robot executes the target landing point sequence and completes its operation.

[0006] In one embodiment, the step of acquiring a discrete map having topographical information corresponding to a walking target area and generating an initial landing point sequence based on the topographical points includes: acquiring a discrete map having topographical information corresponding to the target walking area, and performing a gradient calculation on the discrete map; determining, as passable points, topographical points in the discrete map where the result of gradient calculation is equal to or less than the maximum landing height of the biped walking robot, and generating a first topographical map based on the passable points; and determining, based on the first topographical map, a sequence of initial landing points for the bipedal robot to pass through the target walking area.

[0007] In one embodiment, the step of acquiring a discrete map having topographical information corresponding to a walking target area and generating an initial landing point sequence based on the topographical points further comprises: The method includes a step of determining a topographical point in the discrete map where the result of gradient calculation is zero as a landing point for the bipedal robot, and updating the first topographical map based on the landing point.

[0008] In one embodiment, the method includes the steps of: determining, in the discrete map, a terrain point where a result of gradient calculation is greater than the maximum landing height as an obstacle point that cannot be passed by the bipedal walking robot; The method further includes a step of assigning grid weights based on the topographical information of the obstacle points and the passable points, and constructing a two-dimensional second topographical map.

[0009] In one embodiment, the step of generating an initial movement trajectory of the biped robot based on the discrete map includes: obtaining a kinematic model of the wheeled robot; acquiring an initial position and posture and a target position and posture of the biped walking robot; The method includes a step of combining the second topographical map, the initial position and posture, and the target position and posture to perform a trajectory planning process based on a kinematic model of the wheeled robot, and generating an initial movement trajectory of the bipedal robot.

[0010] In one embodiment, the method further comprises determining optimized sizing parameters and ranges for the biped robot and establishing a kinematic model of the biped robot.

[0011] In one embodiment, the step of determining optimized sizing parameters and ranges for the biped robot and establishing a kinematic model of the biped robot comprises: Obtaining at least one of a motion step length, a lateral movement distance, and a body orientation angle of the biped walking robot; and optimizing positions of landing points of two supporting legs of the bipedal robot based on constraint relationships related to at least one of the movement step length, the lateral movement distance, and the trunk orientation angle.

[0012] In one embodiment, the step of optimizing the positions of the landing points of the two supporting legs of the bipedal robot based on a constraint relationship related to at least one of the movement step length, the lateral movement distance, and the trunk orientation angle includes: optimizing the positions of the landing points of the two supporting legs of the bipedal walking robot so that the positional relationship between the two landing points satisfies the constraint relationship between the motion step length and the landing point range; optimizing the positions of the landing points of the two supporting legs of the bipedal robot so that the positional relationship between the two landing points matches the lateral movement distance; optimizing the plantar orientation angles of the two supporting feet of the bipedal robot based on a predetermined magnification so that the ratio of the sum of the orientation angles of the two plantars to the trunk orientation angle reaches a predetermined magnification.

[0013] The present invention further provides a motion planning device for a bipedal robot, the motion planning device for the bipedal robot comprising: a motion constraint module for acquiring a discrete map having topographical information corresponding to a target walking area, determining topographical points in the discrete map that the bipedal walking robot can pass through, and generating an initial landing point sequence based on the topographical points; a trajectory generation module for generating an initial motion trajectory of the biped walking robot based on the discrete map; and a sequence generation module for optimizing the initial landing point sequence according to the initial motion trajectory and a kinematic model of the bipedal robot based on a predetermined objective function to generate a target landing point sequence so that the bipedal robot executes the target landing point sequence and completes its movement.

[0014] The present invention further provides a motion planning device for a bipedal robot, the motion planning device for a bipedal robot including a memory, a processor, and a bipedal robot motion planning program stored in the memory and executable on the processor, and when the processor executes the bipedal robot motion planning program, the device performs the steps of the above-mentioned bipedal robot motion planning method.

[0015] The present invention further provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the above-described motion planning method for a bipedal robot.

[0016] According to an embodiment of the present invention, a motion planning method, apparatus, device, and storage medium for a bipedal robot generate an initial landing point sequence based on terrain points passable by the bipedal robot in a discrete map having terrain information corresponding to a target area, generate an initial motion trajectory for the bipedal robot based on the discrete map, and optimize the initial landing point sequence in accordance with the initial motion trajectory and a kinematic model of the bipedal robot based on a predetermined objective function to generate a target landing point sequence, so that the bipedal robot executes the target landing point sequence by combining terrain characteristics to more efficiently perform the task of passing through the target area. Here, the kinematic characteristics of the bipedal robot are used to perform a discretization process on terrain with differences in elevation, and parameters of the bipedal robot's landing points are optimized based on ground height gradient information in the discrete map (including determining whether terrain points are passable by the bipedal robot or can be used as landing points for the bipedal robot), thereby obtaining the target landing point sequence.

[0017] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. [Brief explanation of the drawings]

[0018] [Figure 1a] 1 is a flowchart of a motion planning method for a biped robot according to an embodiment of the present invention. [Figure 1b] 10 is a flowchart of a motion planning method for a biped robot according to another embodiment of the present invention. [Figure 2] 1 is a schematic diagram illustrating the movement of a bipedal robot according to an embodiment of the present invention. [Figure 3a]FIG. 1 is a schematic diagram illustrating terrain discretization according to an embodiment of the present invention. [Figure 3b] FIG. 1 is a schematic diagram illustrating terrain discretization according to an embodiment of the present invention. [Figure 3c] FIG. 1 is a schematic diagram illustrating terrain discretization according to an embodiment of the present invention. [Figure 4] FIG. 2 is a schematic diagram showing a plan view of landing point constraints for a bipedal robot according to an embodiment of the present invention. [Figure 5a] FIG. 1 is a schematic diagram illustrating discrete landing point planning according to an embodiment of the present invention. [Figure 5b] FIG. 1 is a schematic diagram illustrating discrete landing point planning according to an embodiment of the present invention. [Figure 5c] FIG. 1 is a schematic diagram illustrating discrete landing point planning according to an embodiment of the present invention. [Figure 6] FIG. 10 is a schematic diagram illustrating planning of discrete landing points according to another embodiment of the present invention. [Figure 7] 1 is a schematic diagram illustrating a motion planning device for a biped walking robot according to an embodiment of the present invention. [Figure 8] FIG. 1 is a schematic diagram illustrating a motion planning device for a biped robot according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0019] In order to make the objectives, technical solutions and advantages of the present invention clearer, the following will clearly and completely describe the technical solutions of the present invention in conjunction with the accompanying drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments that can be obtained by those skilled in the art without any creative efforts belong to the protection scope of the present invention.

[0020] Hereinafter, embodiments of the present invention will be described in detail.

[0021] 1 is a flowchart of a method according to an exemplary embodiment of the present invention. As shown in FIG. 1, the method includes the following steps 110 to 130.

[0022] In step 110, a discrete map having terrain information corresponding to the target walking area is obtained, terrain points in the discrete map that can be passed by the bipedal walking robot are determined, and an initial landing point sequence is generated based on the terrain points.

[0023] In step 120, an initial movement trajectory of the biped robot is generated based on the discrete map.

[0024] In step 130, the initial landing point sequence is optimized according to the initial motion trajectory and the kinematic model of the bipedal robot based on a predetermined objective function to generate a target landing point sequence, so that the bipedal robot executes the target landing point sequence and completes its operation.

[0025] This embodiment aims to generate a series of landing points by utilizing known terrain information and the kinematic characteristics of a bipedal robot (legged robots can traverse terrain of different heights), and to realize a movement task in which the bipedal robot moves from a specified position and posture through the terrain to a target position and posture. Here, "pose" is a general term for position and orientation, and can be divided into two-dimensional and three-dimensional positions and postures. Two-dimensional positions and postures include information on a total of three degrees of freedom, namely, horizontal position and vertical orientation. Three-dimensional positions and postures include information on a total of six degrees of freedom, namely, three-dimensional position and orientation. In this invention, the positions and postures are two-dimensional positions and postures, and specifically include position information of the landing points of the bipedal robot and information on the orientation of the torso.

[0026] For example, the motion planning method for a bipedal robot can be applied to a motion planning system for a bipedal robot, and the system can be applied to a motion planning device for a bipedal robot. The above steps will be described in detail below.

[0027] In step 110, a discrete map having terrain information corresponding to the target walking area is obtained, terrain points in the discrete map that can be passed by the bipedal walking robot are determined, and an initial landing point sequence is generated based on the terrain points.

[0028] Service robots can be broadly categorized into wheeled and legged robots based on their structure. Wheeled robots use the rotation of the wheels on their chassis as their main driving force. Compared to wheeled robots, legged robots have higher mobility and can navigate terrain with large differences in elevation (such as stairs). As shown in Figure 2, a bipedal robot can ascend stairs from position A and descend stairs to reach position B, enabling it to handle a wider range of walking scenarios than wheeled robots. However, legged robots have a complex structure and the system itself has many unstable elements. Bipedal robots, in particular, must consider relevant constraints in the trajectory planning process to ensure the safe execution of the entire mobility task.

[0029] The kinematic characteristics of the bipedal robot are utilized to perform discretization processing on terrain with differences in elevation, and the terrain characteristics are combined to more efficiently complete the traversal task. Specifically, a discrete map is obtained based on known terrain information of the area to be traversed by the bipedal robot, and based on the ground height gradient information in the discrete map, it is determined whether the location can be traversed by the bipedal robot or whether it can be used as a landing point for the bipedal robot.

[0030] For example, in a discrete map, if the height of a terrain change is greater than the maximum foot-lift height of the bipedal robot, it indicates that the bipedal robot cannot pass through that location directly, and conversely, if the height of the terrain change is equal to or less than the maximum foot-lift height of the bipedal robot, it indicates that the bipedal robot can pass through that location. In order to clearly explain the kinematic characteristics of a bipedal robot passing through different terrain heights, the following will use stairs as an example to describe the movement situations of a bipedal robot when facing different terrain changes.

[0031] For example, referring to Figure 3a, if the terrain height range is concentrated in the (0, 0.10, 0.20) range, the terrain height in region C is 0 meters, the terrain height in region D is 0.10 meters, and the terrain height in region E is 0.20 meters, and the maximum foot-lift height of a bipedal robot is 0.15 meters. Clearly, a bipedal robot cannot directly lift its feet from the ground in region C (with a terrain height of 0 meters) to the ground in region E (with a terrain height of 0.20 meters, where steps are formed relative to region C). Therefore, the terrain discretization method marks the boundary point between the two terrains, C and E, which cannot be directly overcome, as an obstacle point. However, a bipedal robot can directly move from the ground in region C (with a terrain height of 0 meters) to the ground in region D (with a terrain height of 0.10 meters, where steps are formed relative to region C). The terrain discretization method marks the boundary point between the two terrains, C and D, which can be directly moved to, as a passable point. The obstacle points and passable points are used in the initial motion trajectory planning process. That is, the passable points form an initial landing point range, and then landing points that can be used for optimization are determined based on the kinematic constraints of the bipedal robot and the landing point range, thereby obtaining an initial landing point sequence. Specifically, the plantar orientations of two adjacent landing points of the bipedal robot can be constrained, that is, the orientation angles of the two adjacent landing points and their average value must not deviate significantly from the orientation angle of the body, so as to reduce the probability of joint motor failure due to the mechanical structure.

[0032] As an example, step 110 specifically includes step 111 of obtaining a discrete map having topographical information corresponding to the target walking area and performing gradient calculations on the discrete map; step 112 of determining, as passable points, topographical points on the discrete map where the result of the gradient calculation is less than or equal to the maximum landing height of the bipedal robot, and generating a first topographical map based on the passable points; and step 113 of determining, based on the first topographical map, a series of initial landing points for the bipedal robot to pass through the target walking area.

[0033] The current maximum foot-lift height h of a biped robot for passing through the target walking areamax Maximum leg lift height h max represents the range of height variations that the bipedal robot can overcome. The maximum foot lift height may be the actual foot lift height of the bipedal robot, or may be a maximum foot lift height specified based on actual needs and safety considerations. As can be understood, the actual foot lift height is between 0 meters and the maximum foot lift height.

[0034] A terrain height map h(x,y) showing the walking area is obtained. Here, x and y represent horizontal position coordinates, and h represents the terrain height at the corresponding horizontal position coordinates. The actual map is gridded to obtain the corresponding discrete map h(i,j), and the gradient of each pixel point (i.e., terrain point) on the discrete map h(i,j) is calculated. If the gradient is equal to the maximum foot lift height h, max Smaller terrain points are marked as passable points, indicating that the biped robot can pass through these passable points, thereby forming a range of initial landing points.

[0035] The passable points determined in this manner form a first topographical map. The passable points in the first topographical map can be candidate landing points for the bipedal robot's motion, and are combined with kinematic constraints on the plantar orientations of two adjacent landing points of the bipedal robot to generate an initial landing point sequence. The initial landing point sequence must be combined with the bipedal robot's path from the initial position and posture to the target position and posture to select an optimal target landing point sequence.

[0036] Step 110 may further include step 112a of determining a topographical point in the discrete map where the result of gradient calculation is zero as a landing point for the bipedal robot, and updating the first topographical map based on the landing point. Here, step 112a may be performed in parallel with step 112, or may be performed before or after step 112, as long as it is performed before step 113.

[0037] In the discrete map, the gradient is the maximum foot height h maxThe smaller passable points include boundary points and non-boundary points between two terrain features. For a bipedal robot, landing on the boundary point of a staircase can easily lead to dangerous situations such as unbalance. Therefore, considering the stability of a bipedal robot, all boundary points are marked as non-landing points.

[0038] For example, terrain points with a non-zero gradient are represented as boundary points, i.e., marked as impossible landing points. Furthermore, terrain points with a zero gradient are represented as non-boundary points and marked as landing points. As shown in Figures 3a to 3c, Figure 3a shows an original discrete map with terrain height information, where the terrain height of area C is 0 meters, the terrain height of area D is 0.10 meters, and the terrain height of area E is 0.20 meters. Through gradient calculation, the area surrounded by the dashed line in Figure 3b is determined as the boundary area. When optimizing the landing point, the landing point cannot be located within the boundary area. Of course, the landing point can be located in an area other than the boundary area in Figure 3b.

[0039] For the first topographical map determined as described above, passable points on the first topographical map are used as candidate landing points for the bipedal robot's motion, and are combined with kinematic constraints on the plantar orientations of two adjacent landing points of the bipedal robot to generate an initial landing point sequence, which must be combined with the bipedal robot's path from the initial position to the target position to select an optimal target landing point sequence.

[0040] In step 120, an initial movement trajectory of the biped robot is generated based on the discrete map.

[0041] Based on the magnitude relationship between the gradient calculated from the discrete map and the maximum foot-lift height of the bipedal robot, the terrain points in the discrete map are marked as obstacle points, landing-impossible points, passable points, etc., to obtain a second terrain map and a first terrain map. Here, the second terrain map includes the marked obstacle points, which can be used as parameters when planning a path.

[0042] As an example, the method may further include a step of determining terrain points in the discrete map where the result of gradient calculation is greater than the maximum landing height as obstacle points that the bipedal robot cannot pass through, and a step of assigning grid weights based on terrain information of the obstacle points and the passable points to construct a two-dimensional second terrain map for subsequent trajectory planning.

[0043] The gradient of each pixel point (i.e., terrain point) in the discrete map is calculated, and the calculated gradient is the maximum foot lift height h of the bipedal robot. max Points with larger gradients are marked as obstacle points. Still referring to Figures 3a-3c, by comparing the magnitude of the gradients of each terrain point in the discrete map, the obstacle points in the area enclosed by the dashed curve shown in Figure 3c are obtained. These obstacle points represent locations that are difficult for the bipedal robot to overcome and need to be avoided in subsequent trajectory planning.

[0044] According to the above embodiment, by combining the relevant parameters of the bipedal robot with the topographical information of the target walking area, the position points or position areas that the bipedal robot can pass through can be determined as an initial landing point series, and then the optimal landing point can be selected from the initial landing point series to generate a target landing point series.

[0045] As an example, step 120 may specifically include step 121 of acquiring a kinematic model of the wheeled robot, step 122 of acquiring an initial position and posture and a target position and posture of the bipedal robot, and step 123 of combining the second topographic map, the initial position and posture and the target position and posture, performing a trajectory planning process based on the kinematic model of the wheeled robot, and generating an initial movement trajectory of the bipedal robot.

[0046] Since the trajectory algorithm for wheeled robots is relatively mature and widely used, the overall trajectory planning problem for legged robots can be solved to some extent by using the trajectory planning method for wheeled robots. For example, the trajectory planning method for wheeled robots can be used to generate a reference trajectory for legged robots. Specifically, the gradient of each terrain grid is calculated using a discrete map, and if the gradient is too large (the gradient is too large for the maximum leg lift height h), max The map grids with a moderate gradient (with a maximum foot lift height h) are set as obstacles that the biped robot cannot pass through. max The map grids (those with a gradient smaller than or equal to zero) are set as passable points for the bipedal robot, and appropriate grid weights are set according to the magnitude of the gradient to construct a two-dimensional second terrain map. The grid weights corresponding to the map grids are penalty factors. For example, in terrain areas that are difficult to cross, the grid weights of the corresponding map grids may be set high. Typically, the grid weights are either 0 or 1, with 1 representing an impassable point and 0 representing a passable point.

[0047] The body motion model of a bipedal robot is abstracted and converted into a virtual wheeled robot. Here, the position and posture information is information for planning the overall motion tendency of the bipedal robot, and includes the position information of the landing point on the second topographical map and the orientation information of the body. The kinematic model of the wheeled robot is expressed as follows:

number

number

[0048] Then, using the above-mentioned kinematic model of the wheeled robot, in combination with differential dynamic programming (DDP), a sufficiently smooth position and posture curve connecting the initial position and posture and the target position and posture is generated.

[0049] In the process of combining the second topographical map, the initial position and posture, and the target position and posture, and performing a trajectory planning process based on the kinematic model of the wheeled robot to generate an initial motion trajectory for the bipedal robot, relevant parameters of the wheeled robot may be set according to the motion parameters of the bipedal robot. For example, an upper limit on the linear velocity of the wheeled robot's motion may be set based on the maximum (forward or backward) speed of the bipedal robot, and an upper limit on the angular velocity of the wheeled robot's motion may be set based on the maximum turning speed of the bipedal robot. As a result, the trajectory constructed based on the virtual wheeled robot model after parameter setting can be used as a reference for the center of gravity movement and torso orientation of the bipedal robot, i.e., as the initial motion trajectory of the bipedal robot for subsequent landing point optimization.

[0050] In step 130, the initial landing point sequence is optimized according to the initial motion trajectory and the kinematic model of the bipedal robot based on a predetermined objective function to generate a target landing point sequence, so that the bipedal robot executes the target landing point sequence and completes its operation.

[0051] The initial motion trajectory is used as a reference trajectory for the biped robot to reach the target position and posture from the initial position and posture, and after defining the kinematic constraints of the biped robot, different target landing point sequences can be obtained based on different optimization goals, so that the biped robot can walk from the starting point to the target point by executing the target landing point sequences.

[0052] Research has shown that bipedal robots and wheeled robots share common kinematic characteristics. For example, without considering wheel slippage, the instantaneous linear velocity of a wheeled robot coincides with the robot's orientation. Similarly, the locomotion ability of a bipedal robot is strongest in the direction the robot is facing. However, because the actual motion of a bipedal robot is discrete, it is necessary to adjust the movement of its lateral trunk according to the landing point at a given time. Furthermore, to ensure the stability of the entire gait, the movement distance of the lateral trunk of a bipedal robot must be kept within a certain range. Therefore, as shown in FIG. 1b, before optimizing the initial landing point sequence according to the initial motion trajectory and the kinematic model of the bipedal robot based on a predetermined objective function to generate a target landing point sequence, the method may further include step 130a of determining optimized dimension parameters and ranges for the bipedal robot and establishing a kinematic model of the bipedal robot.

[0053] By establishing a kinematic model of the bipedal robot based on the structure, motion parameters, joint activity parameters, etc. of the bipedal robot, the range of motion of the bipedal robot can be effectively restricted, thereby ensuring that the bipedal robot can stably pass through areas with changing terrain.

[0054] As an example, step 130a may specifically include the steps of: obtaining at least one of a movement step length, a lateral movement distance, and a trunk orientation angle of the bipedal robot; and optimizing positions of landing points of two supporting legs of the bipedal robot based on constraint relationships related to at least one of the movement step length, the lateral movement distance, and the trunk orientation angle. Here, the step of optimizing the positions of the landing points of the two supporting feet of the bipedal robot based on a constraint relationship related to at least one of the movement step length, the lateral movement distance, and the torso orientation angle may include at least one of the following steps: optimizing the positions of the landing points of the two supporting feet of the bipedal robot so that the positional relationship between the two landing points conforms to the constraint relationship between the movement step length and the landing point range; optimizing the positions of the landing points of the two supporting feet of the bipedal robot so that the positional relationship between the two landing points conforms to the lateral movement distance; and optimizing the orientation angles of the soles of the two supporting feet of the bipedal robot based on a predetermined multiplier so that the ratio of the sum of the orientation angles of the two soles of the feet to the torso orientation angle reaches a predetermined multiplier.

[0055] Specifically, it is understood that the distance that a biped walking robot moves forward or backward must be kept within a certain range. Therefore, the present invention proposes a kinematic constraint for the displacement between adjacent landing points of a biped walking robot. For example, referring to FIG. 4, the positions and postures of two adjacent landing points of a biped walking robot are respectively expressed as (x p ,y p ,θ p ), (x n ,y n ,θ n ) and the displacement between two adjacent landing points must satisfy the following relation (1):

number

[0056] It is understood that the lateral stride of a bipedal robot must not deviate significantly from shoulder width, otherwise unnecessary lateral vibration of the trunk will occur, reducing the stability of walking. Therefore, the present invention proposes a kinematic constraint on the lateral movement distance of a bipedal robot, i.e., the distance between the two landing points in front and behind the bipedal robot in a direction perpendicular to the direction of the trunk should be kept close to shoulder width. For example, referring to FIG. 4, the positions and postures of two adjacent landing points of a bipedal robot are respectively (x p ,y p ,θ p ), (x n ,y n ,θ n ) and the lateral movement distance of a bipedal robot must satisfy the following relational expression (2).

number

[0057] It is also understood that the orientation angles of two adjacent landing points and their average value must not deviate significantly from the orientation angle of the body. Therefore, the present invention proposes a kinematic constraint on the orientation angle between adjacent landing points of a biped robot. For example, referring to FIG. 4, the positions and orientations of two adjacent landing points of a biped robot are respectively (x p ,y p ,θ p ), (x n ,y n ,θ n ) and the directions of the soles of the feet at the two landing points before and after must satisfy the following relational expression (3).

number

[0058] The above constraints can effectively reduce the search space for landing points of the bipedal robot, thereby improving the efficiency of subsequent optimization of landing points. For example, as shown in Figure 4, the dashed area F in which the right supporting foot is located is the possible landing range for the right supporting foot.

[0059] 5a to 5c, a bipedal robot must reach a target position / posture B from an initial position / posture A in FIG. 5a by passing through three regions of different terrain heights on the CDE. An optimal landing point sequence is generated based on the kinematic model of the bipedal robot described above and the initial motion trajectory shown in FIG. 5b. The initial values ​​of the landing point sequence are the aforementioned initial landing point sequence, which includes the number of landing points, the horizontal position coordinates of each landing point, their orientation, and their phase (whether they correspond to left-foot support or right-foot support), an example of which is shown in FIG. 5c. Therefore, to generate the landing point sequence, the following optimization problem is formulated:

number

number

number

number

number

number

[0060] Under the assumption that the task execution time for each landing point is constant, the overall task execution time for the complete sequence of landing points is proportional to the number of landing points. Therefore, a larger value of the time weight λ0 indicates that the task requires the shortest possible execution time, while a smaller value of λ0 indicates that the task is less sensitive to its execution time. Also, a λ0 of 0 means that the task does not care about the execution time at all.

[0061] In addition, the state of the landing point also affects the success rate of task execution and the safety of the robot, so the sensitivity of a specific task to different landing point states is different. Generally, the first and last steps of a bipedal robot correspond to the start and stop stages of its movement, and these two stages involve the accumulation and dissipation of the entire body momentum, which is likely to cause unstable situations. Therefore, if the stability requirement of the task being performed is high, the state weights λ1 and λ2 of the first and last landing points should be adjusted accordingly. n Increase the value of , and the state weights λ of the remaining intermediate destinations i(i=2,3,...,n-1) may use approximately the same value.

[0062] Since the landing point of a bipedal robot is related to the two supporting legs (left and right), the present invention proposes a concept of the phase of the landing point, which is exemplarily represented by the following equation (4).

number

[0063] The present invention also proposes a kinematic constraint on the phase of two adjacent landing points, which is expressed as the following equation (5):

number

[0064] As mentioned above, to avoid the risk of imbalance associated with a bipedal robot stepping onto a terrain point with a non-zero gradient, the landing point of the bipedal robot should be a terrain point with a zero gradient, specifically expressed by the following equation (6):

number

[0065] Based on the above equations (1) to (3), (x p ,y p ) is the horizontal coordinate of the current landing point, and (x n ,y n ) is the horizontal position coordinate of the next landing point, and using i to identify the current landing point and i+1 to identify the next landing point, the kinematic constraints of a biped robot can be summarized as follows:

number

number

[0066] When generating a sequence of target landing points using the above optimization problem, the initial and final conditions must be satisfied, as shown in Figure 2 and Figure 5a. That is, the position and orientation of the first landing point must match the initial position and orientation, and the position and orientation of the last landing point must match the target position and orientation.

[0067] Furthermore, the weighting parameters in the optimization objective may be selected according to the actual task scenario. For example, if a bipedal robot is required to reach a target position and posture in the shortest time, the number n of landing point sequences may be used as the target parameter. Alternatively, if a bipedal robot is required to move to a target position as smoothly as possible, the sum of the weights of the landing points on a discrete topographical map may be used as the target parameter. In this way, by rationally adjusting the weights of the objective function, the task execution performance of a legged robot can be effectively improved.

[0068] In the following, an example is shown in which the sum of the weights of the landing points on the discrete topographical map is used as the target parameter.

[0069] λ i=1,2,...,nWhen is zero, the overall objective function is the number of landing points, i.e., the bipedal robot needs to be controlled so that it can reach the target position and posture from the initial position and posture in the fewest number of steps. For example, as shown in Figures 5a to 5c, the execution path of the bipedal robot is "turn left - go up the stairs between areas C and D - move straight - go down the stairs between areas D and C - turn right - reach the final position and posture," i.e., it needs to move by making full use of the terrain.

[0070] If the overall objective function is the height of the landing sequence, the corresponding optimal solution is to minimize the number of stair climbing steps. For example, as shown in Figures 5a to 5c, the planned landing does not have any steps to climb or descend stairs, and the left foot moves on the flat ground and the right foot moves on the stairs until the specified position is reached from the initial position.

[0071] In another embodiment, with reference to Figure 6, the relevant kinematic constraints for a biped robot are as follows: Based on the structure and kinematic characteristics of the biped robot, the maximum forward step length d f is 0.5 meters, maximum backward step length d b is 0.3 meters, the distance between the two support legs h is 0.5 meters, and the maximum allowable lateral deviation distance d s is 0.05 meters, and the maximum rotation angle of the support foot is θ max is 10 degrees.

[0072] Initial position (0,0,0) and target position

number

number

[0073] Still referring to FIG. 6, it is assumed that the prohibited area for landing points obtained by terrain discretization can be represented as the gray area G[0.6,0.7]×[0,0.6] in FIG.

[0074] Based on plane geometry knowledge, if the positions (x1, y1) and (x2, y2) of the two landing points before and after are given, the angle at which the line connecting them intersects with the reference trajectory can be calculated using the following analytical formula.

number

[0075] Assuming that the trajectory length of the reference trajectory is L, the fixed step length of each step is L / n, and the average value of the plantar orientation angles of two adjacent landing points of a bipedal robot is equal to the trunk orientation angle of the bipedal robot, the initial landing point sequence in the above embodiment can be obtained by the following equation:

number

[0076] The initial landing point sequence calculated using the above formula does not necessarily satisfy the final boundary condition. It is necessary to gradually determine a target landing point sequence that satisfies the final boundary condition using the boundary value algorithm of IQP (Iterative Quadratic Programming) by combining the kinematic constraints of the bipedal robot, the initial motion trajectory, and a predetermined objective function.

[0077] Specifically, assuming that the number of landing points n is 2, the parameters that need to be optimized are the position, posture, and support phase (x1, y1, θ1, δ1) of the first landing point, and the position, posture, and support phase (x2, y2, θ2, δ2) of the second landing point. Substituting the above parameters into the above constraints (1) to (6), the corresponding relational expressions are expressed as follows:

number

[0078] Based on the above constraints, it can be determined that the quantities that need to be optimized are the support phase of the first landing point and the positions of the two landing points, and all other relevant quantities can be calculated based on the above constraints. In some embodiments, an objective function expressed as follows may be used:

number

[0079] The physical meaning of this objective function is to minimize the difference in plantar orientation between the two intermediate landing points. This is because, in actual applications, if the difference in rotation angle between the two supporting legs of a bipedal robot is too large, some of the joint motors will be prone to approaching mechanical limits, and self-collisions will be more likely to occur during the leg swing process. This objective function can significantly reduce the probability of the above problems occurring and play a role in protecting the robot.

[0080] In some embodiments, the following objective function may be used:

number

[0081] The physical meaning of this objective function is to reduce the degree of lateral deviation between adjacent landing points. This is because, during the movement of a bipedal robot, lateral deviation often causes the body to sway, and if the center of gravity of the body is too high, it becomes unstable and the robot will fall to the side. This objective function can effectively reduce the width of lateral deviation and improve the stability of the movement.

[0082] In some embodiments, an objective function for the number n of landing points may be used. Specifically, the motion of a bipedal robot is often assumed to have the same execution time for each step, and minimizing the number n of landing points means the fewest number of steps, i.e., the shortest motion time. In other words, the physical meaning of the objective function of minimizing the number n of landing points is to reduce the time required for the bipedal robot to move from the initial position and posture to the target position and posture, i.e., it is required to fully utilize the robot's motor capabilities to reach the target position and posture from the initial position and posture in the shortest time.

[0083] A landing point sequence optimization problem is solved based on the objective function to generate a target landing point sequence. Briefly, when the number of landing points is determined (e.g., when n = 2 as described above), the phase value of the supporting foot corresponding to the first landing point is either 0 or 1, and these two cases may be calculated separately. In particular, when the phase value of the supporting foot is determined, the kinematic constraints of the bipedal robot described above become nonlinear constraint vectors, and the corresponding objective function is usually nonlinear. Therefore, a local optimum can be obtained by solving the problem using the iterative quadratic programming (IQP) method. Then, optimal target values ​​are calculated for the two cases in which the phase of the supporting foot corresponding to the first landing point is 0 or 1, and the supporting foot phase with the smaller optimal target value is designated as the final solution.

[0084] However, in practice, the number of landing points is usually uncertain, so it is necessary to search for the number of landing points that can satisfy all constraints. This can also be thought of as searching for the number of landing points for which a solution exists. Furthermore, from the perspective of search efficiency, a binary search can be performed on the number of landing points to find a solution in a relatively small number of iterations, assuming that an optimal solution exists. For example, since a landing point sequence must include at least the first landing point (i.e., the initial position and posture) and the last landing point (i.e., the target position and posture), the lower limit of the number of landing points is clearly two. Therefore, it is possible to first check whether a solution exists when there are two landing points. If a solution exists when there are two landing points, it means that the bipedal robot can directly reach the target position and posture from the initial position and posture (which is basically impossible). If a solution does not exist when there are two landing points, the number of landing points is doubled to obtain a new landing point number of four, and the process of checking whether a solution exists for the new number of landing points is continued. If a solution does not exist even with the new number of landing points, a new number of landing points is generated and the search continues. This process is repeated until, if a solution exists, a corresponding number of landing points is found and determined to be the smallest number of landing points that can satisfy the overall constraints, or until the number of new landing points generated is too large (e.g., exceeds a predetermined threshold) and the process of searching for that number of landing points is terminated.

[0085] Alternatively, if it is predetermined that there exists a certain number of landing points n that can satisfy the overall constraints, a new number of landing points (n+1) / 2 can be selected and a check is made to see if a solution exists at this time. If a solution exists for the new number of landing points (n+1) / 2, a new number of landing points (n+1) / 4 is again generated and a check is made to see if a solution exists at this time. If a solution exists for the new number of landing points (n+1) / 4, a new number of landing points is again generated and the search continues. This process is repeated until a landing point number corresponding to a case where no solution exists is found and the previous number of landing points corresponding to that landing point number is determined to be the smallest number of landing points that can satisfy the existence of a solution, i.e., the smallest number of landing points that satisfies the overall constraints.

[0086] Although the above describes a method for searching for the number of landing points using a bisection method as an example, those skilled in the art should understand that the bisection method is used merely to consider search efficiency and that the present invention should not be limited thereto. In fact, any other search method known to those skilled in the art (e.g., a method for generating a new number of landing points by gradually increasing the number of landing points from a lower limit value for the number of landing points at a specific step length, or a method for generating a new number of landing points by gradually decreasing the number of landing points from the number of landing points at which a predicted solution exists at a specific step length) can also be applied to embodiments of the present invention. Furthermore, if the objective function includes the number of landing points and other related parameters, the minimum value of the objective function may be recorded throughout the process of searching for the number of landing points, and the minimum value of the objective function and the corresponding solution may be reported when the minimum number of landing points is found.

[0087] A bipedal robot motion planning method according to the present invention generates an initial landing point sequence based on terrain points passable by the bipedal robot in a discrete map having terrain information corresponding to a target area, generates an initial motion trajectory for the bipedal robot based on the discrete map, and optimizes the initial landing point sequence in accordance with the initial motion trajectory and a kinematic model of the bipedal robot based on a predetermined objective function to generate a target landing point sequence, so that the bipedal robot executes the target landing point sequence and combines terrain characteristics to more efficiently perform the task of passing through the target area. Here, the kinematic characteristics of the bipedal robot are used to perform a discretization process on terrain with differences in elevation, and parameters of the bipedal robot's landing points are optimized based on ground height gradient information in the discrete map (including determining whether terrain points are passable by the bipedal robot or can be used as landing points for the bipedal robot), thereby obtaining the target landing point sequence.

[0088] Based on the same idea as the above-mentioned method, the present invention provides a motion planning device for a biped walking robot, as shown in Figure 7. The motion planning device for the biped walking robot comprises: a motion constraint module 702 for acquiring a discrete map having terrain information corresponding to a target walking area, determining terrain points in the discrete map that the bipedal walking robot can pass through, and generating an initial landing point sequence based on the terrain points; a trajectory generation module 704 for generating an initial movement trajectory of the biped robot based on the discrete map; and a sequence generation module 706 for optimizing the initial landing point sequence according to the initial motion trajectory and a kinematic model of the bipedal robot based on a predetermined objective function to generate a target landing point sequence so that the bipedal robot executes the target landing point sequence and completes its operation.

[0089] Optionally, the motion constraint module 702: acquiring a discrete map having topographical information corresponding to the walking target area, and performing a gradient calculation on the discrete map; determining, as passable points, terrain points in the discrete map where the result of gradient calculation is equal to or less than the maximum landing height of the biped walking robot, and generating a first topographical map based on the passable points; Based on the first topographical map, a sequence of initial landing points for the bipedal robot to pass through the target walking area is determined.

[0090] Optionally, the motion constraint module 702: A topographical point in the discrete map where the result of gradient calculation is zero is set as a landing point for the bipedal robot, and is used to update the first topographical map based on the landing point.

[0091] Optionally, the motion constraint module 702: a topographical point in the discrete map where the result of gradient calculation is greater than the maximum landing height is determined as an obstacle point that the bipedal walking robot cannot pass through; Grid weights are assigned based on the topographical information of the obstacle points and the passable points and are used to construct a two-dimensional second topographical map.

[0092] Optionally, the trajectory generation module 704: Obtain the kinematic model of the wheeled robot, Acquire an initial position and posture and a target position and posture of the biped walking robot; The second topographic map, the initial position and posture, and the target position and posture are combined to perform a trajectory planning process based on a kinematic model of the wheeled robot, which is used to generate an initial movement trajectory of the bipedal robot.

[0093] Optionally, the sequence generation module 706: The optimized size parameters and ranges of the biped robot are determined and used to establish a kinematic model of the biped robot.

[0094] Optionally, the sequence generation module 706: acquiring at least one of a motion step length, a lateral movement distance, and a body orientation angle of the biped walking robot; The method is used to optimize the positions of the landing points of the two supporting legs of the bipedal robot based on a constraint relationship related to at least one of the movement step length, the lateral movement distance, and the trunk orientation angle.

[0095] The process for realizing the functions and roles of each module / sub-module / unit in the above device is specifically detailed in the process for realizing the corresponding step in the above method, and similar technical effects can be obtained, so it will not be repeated here.

[0096] Corresponding to the above method embodiment, the present invention further provides an apparatus and a terminal to which the apparatus is applied.

[0097] An embodiment of the bipedal robot motion planning device of the present invention may be applied to a computer device such as a server or a terminal device. The device may be implemented using software, hardware, or a combination of software and hardware. For example, in the case of software implementation, the device is formed by loading computer program instructions corresponding to non-volatile memory into an internal memory and executing them via a bipedal robot motion planning processor in which the instructions are implemented. At the hardware level, FIG. 8 is a hardware structural diagram of a computer device in which a bipedal robot motion planning device 831 according to an embodiment of the present invention is implemented. In addition to the processor 810, internal memory 830, network interface 820, and non-volatile memory 840 shown in FIG. 8 , any device with data processing capabilities in which the device is implemented may typically include other hardware depending on the actual functions of the computer device, and further description thereof will be omitted here.

[0098] The foregoing describes specific embodiments of the present invention. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in a different order than in the examples and still achieve desirable results. Also, the procedures depicted in the figures do not necessarily require the particular order or sequential order shown to achieve expected results. In some embodiments, multitasking and parallel processing may also be possible or advantageous.

[0099] Other embodiments of the invention will be readily apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. The present invention is intended to cover any modifications, uses, or adaptations of the invention, which modifications, uses, or adaptations follow the general principles of the invention and include common knowledge or customary techniques known in the art but which are not disclosed herein. The actual scope and spirit of the invention are indicated by the following claims.

[0100] It should be understood that the present invention is not limited to the exact construction described above and illustrated in the accompanying drawings, but that various modifications and changes are possible without departing from the scope thereof, which is limited only by the appended claims.

[0101] The above is only a preferred embodiment of the present invention, and is not intended to limit the present invention. Those skilled in the art will appreciate that the present invention may undergo various modifications and variations. Any modifications, equivalent replacements, improvements, etc. made without departing from the spirit and principle of the present invention shall fall within the scope of protection of the present invention.

Claims

1. acquiring a discrete map having topographical information corresponding to a target walking area, determining topographical points on the discrete map that can be passed by the bipedal walking robot, and generating an initial landing point sequence based on the topographical points; generating an initial motion trajectory of the biped walking robot based on the discrete map; optimizing the initial landing point sequence according to the initial motion trajectory and a kinematic model of the bipedal robot based on a predetermined objective function to generate a target landing point sequence, so that the bipedal robot executes the target landing point sequence and completes its movement; A motion planning method for a bipedal robot.

2. The step of obtaining a discrete map having topographical information corresponding to a walking target area and generating an initial landing point sequence based on the topographical points includes: acquiring a discrete map having topographical information corresponding to the target walking area, and performing a gradient calculation on the discrete map; determining, as passable points, topographical points in the discrete map where the result of gradient calculation is equal to or less than the maximum landing height of the biped walking robot, and generating a first topographical map based on the passable points; determining, based on the first topographical map, a sequence of initial landing points for the bipedal robot to pass through the target walking area; 2. The method for planning motion for a biped robot according to claim 1.

3. The step of acquiring a discrete map having topographical information corresponding to a walking target area and generating an initial landing point sequence based on the topographical points further includes: a step of determining a topographical point in the discrete map where a result of gradient calculation is zero as a landing point of the bipedal walking robot, and updating the first topographical map based on the landing point; 3. The motion planning method for a biped robot according to claim 2.

4. determining, in the discrete map, a topographical point where the result of gradient calculation is greater than the maximum landing height as an obstacle point that cannot be passed by the bipedal walking robot; and constructing a two-dimensional second topographical map by assigning grid weights based on topographical information of the obstacle points and the passable points.

4. The motion planning method for a bipedal robot according to claim 2 or 3.

5. The step of generating an initial movement trajectory of the biped walking robot based on the discrete map includes: obtaining a kinematic model of the wheeled robot; acquiring an initial position and posture and a target position and posture of the biped walking robot; and performing a trajectory planning process based on a kinematic model of the wheeled robot by combining the second topographical map, the initial position and posture, and the target position and posture to generate an initial motion trajectory of the bipedal robot.

5. The motion planning method for a biped robot according to claim 4.

6. determining optimized sizing parameters and ranges for the biped robot and establishing a kinematic model of the biped robot; 6. The motion planning method for a bipedal robot according to claim 1.

7. The step of determining optimized sizing parameters and ranges of the biped robot and establishing a kinematic model of the biped robot includes: Obtaining at least one of a motion step length, a lateral movement distance, and a body orientation angle of the bipedal robot; optimizing positions of landing points of two supporting legs of the bipedal robot based on constraint relationships related to at least one of the movement step length, the lateral movement distance, and the trunk orientation angle; 7. The method for planning motion for a biped robot according to claim 6.

8. The step of optimizing positions of landing points of two supporting legs of the bipedal robot based on a constraint relationship related to at least one of the movement step length, the lateral movement distance, and the trunk orientation angle includes: optimizing the positions of the landing points of the two supporting legs of the bipedal walking robot so that the positional relationship between the two landing points satisfies the constraint relationship between the motion step length and the landing point range; optimizing positions of landing points of two supporting legs of the bipedal robot so that a positional relationship between the two landing points conforms to the lateral movement distance; optimizing the plantar orientation angles of the two supporting legs of the bipedal robot based on a predetermined scaling factor so that a ratio of the sum of the plantar orientation angles of the two supporting legs to the trunk orientation angle reaches a predetermined scaling factor; 8. The motion planning method for a biped robot according to claim 7.

9. a motion constraint module for acquiring a discrete map having topographical information corresponding to a target walking area, determining topographical points in the discrete map that the bipedal walking robot can pass through, and generating an initial landing point sequence based on the topographical points; a trajectory generation module for generating an initial motion trajectory of the biped walking robot based on the discrete map; a sequence generation module for optimizing the initial landing point sequence according to the initial motion trajectory and a kinematic model of the bipedal robot based on a predetermined objective function to generate a target landing point sequence so that the bipedal robot executes the target landing point sequence and completes an operation; A motion planning device for a bipedal robot.

10. A motion planning device for a bipedal robot, comprising: a memory; a processor; and a motion planning program for a bipedal robot stored in the memory and executable on the processor, wherein when the processor executes the motion planning program for the bipedal robot, the device performs steps of the motion planning method for a bipedal robot according to any one of claims 1 to 8. A motion planning device for a bipedal robot.

11. A computer-readable storage medium storing a motion planning program for a bipedal robot, wherein, when the motion planning program for the bipedal robot is executed, the steps of the motion planning method for a bipedal robot according to any one of claims 1 to 8 are performed. A computer-readable storage medium comprising:

Images