Quadruped robot gait control method and system based on complex terrain

Abstract

The invention discloses a quadruped robot gait control method and system based on complex terrains, and relates to the technical field of robot control. According to the method, based on mechanism parameter characteristics of the quadruped robot, a dynamics and kinematics mathematical model of the quadruped robot and a quantification mode of the terrain unevenness degree are established, and estimation of the foot end contact state probability in the movement process of the robot is achieved; different gait types are designed and matched with the irregular terrains according to robot movement under the irregular terrains, and stability and switching strategies of gaits under different terrains are analyzed; based on the contact state probability of the foot ends, the contact state of the foot ends is judged, real-time optimization of gait phases is achieved, and dynamic balance control over the quadruped robot in the movement process is achieved.

Description

Technical Field

[0001] The present invention relates to the field of robot control technology, and in particular to a gait control method and system for a quadruped robot based on complex terrain. Background Art

[0002] With the rapid development of science and technology, robots have been deeply integrated into various fields. Quadruped robots have become a research focus in the field of robotics due to their excellent terrain adaptability. As a typical nonlinear, multi-input and multi-output drive system, the motion control system of quadruped robots faces technical difficulties in adapting to irregular terrain. Although existing control algorithms perform well on flat ground, when faced with irregular terrain such as discrete contact surfaces and steep slopes, traditional control frameworks show limitations in motion stability and terrain adaptability. When faced with irregular terrain, existing quadruped robot gait control algorithms cannot meet the requirements of online generation of foot trajectories and trunk motion planning that fit the terrain characteristics during motion control, and cannot achieve multi-joint coordinated motion and accurate estimation of the body state during motion, and cannot simultaneously guarantee efficient execution of motion commands and global stability. Summary of the Invention

[0003] The purpose of the present invention is to provide a gait control method and system for a quadruped robot based on complex terrain to improve the above technical problems.

[0004] In order to achieve the above-mentioned object of the invention, the embodiment of the present invention provides the following technical solutions:

[0005] A gait control method for a quadruped robot based on complex terrain, comprising:

[0006] Obtain the mechanism parameter characteristics of the quadruped robot and construct the mechanical motion model of the quadruped robot;

[0007] Simulate the operation of a quadruped robot in complex terrain, and determine the quadruped robot's motion state estimation results, terrain estimation results, and foot end contact estimation results based on the quadruped robot's mechanical motion model;

[0008] Based on the motion state estimation results, terrain estimation results and foot contact estimation results, the gait of the quadruped robot is adjusted in real time.

[0009] A quadruped robot gait control system based on complex terrain, comprising:

[0010] A quadruped robot mechanical motion model construction module is used to obtain the mechanism parameter characteristics of the quadruped robot and construct the quadruped robot mechanical motion model;

[0011] The quadruped robot mechanical motion model simulation module is used to simulate the operation of the quadruped robot in complex terrain and determine the quadruped robot's motion state estimation results, terrain estimation results, and foot end contact estimation results based on the quadruped robot's mechanical motion model;

[0012] The real-time gait adjustment module of the quadruped robot is used to adjust the gait of the quadruped robot in real time based on the motion state estimation results, terrain estimation results and foot end contact estimation results.

[0013] The beneficial effects of the present invention are:

[0014] Based on the parameter characteristics of the quadruped robot mechanism, the present invention establishes its dynamic and kinematic mathematical models and a quantitative method for the degree of terrain unevenness, thereby realizing the estimation of the contact state probability of the foot end during the robot movement; for the robot movement under irregular terrain, different gait types are designed to match the irregular terrain, and the stability and switching strategy of the gait under different terrains are analyzed; the contact state of the foot end is judged based on the contact state probability of the foot end and the gait phase is optimized in real time, thereby realizing dynamic balance control during the movement of the quadruped robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 A flow chart of a method in an embodiment of the present invention;

[0017] Figure 2 Schematic diagram of the support plane of the quadruped robot in an embodiment of the present invention;

[0018] Figure 3 is a graph showing phase progress and first contact probability in an embodiment of the present invention;

[0019] Figure 4 is a graph showing the foot end force and the second contact probability in an embodiment of the present invention;

[0020] Figure 5 is a graph showing the relationship between foot end height and the third contact probability in an embodiment of the present invention;

[0021] Figure 6 Schematic diagram of the gait phase of the quadruped robot in an embodiment of the present invention;

[0022] Figure 7 Schematic diagram of the foot trajectory corresponding to the optimized candidate gait in an embodiment of the present invention;

[0023] Figure 8 A system structure diagram in an embodiment of the present invention;

[0024] Figure 9 is a position curve diagram of the quadruped robot in an embodiment of the present invention;

[0025] Figure 10 : is a speed curve diagram of the quadruped robot in an embodiment of the present invention;

[0026] Figure 11 This is a simulation diagram of a quadruped robot walking on a steep slope in an embodiment of the present invention;

[0027] Figure 12 1 is a curve diagram of the pitch angle change of the quadruped robot when walking on a steep slope in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention.

[0029] See also Figure 1 This embodiment provides a quadruped robot gait control method based on complex terrain, which includes:

[0030] S1. Obtain the structural parameter characteristics of the quadruped robot and establish a mechanical motion model of the quadruped robot.

[0031] Said S1 comprises:

[0032] S1-1. Obtaining mechanism parameter characteristics and setting multiple coordinate systems; the mechanism parameter characteristics include link translation, joint angle, link rotation offset and link extension.

[0033] Among them, the multiple coordinate systems set up are the world coordinate system, the center of mass coordinate system, the fuselage coordinate system, and the leg base coordinate system, hip joint coordinate system, thigh coordinate system and calf coordinate system corresponding to each single leg.

[0034] The quadruped robot's mechanical body consists of a complex motion system consisting of a trunk and distributed leg structures. The left front leg, right front leg, right hind leg, and left hind leg of the quadruped robot are numbered 0, 1, 2, and 3, respectively. The multi-jointed leg structure of the quadruped robot complicates its kinematic description. This embodiment constructs a hierarchical Cartesian coordinate system based on the principles of rigid-body kinematics, thereby establishing a complete spatial motion description framework.

[0035] In the present invention, a world coordinate system is established as a spatial reference. The origin of the world coordinate system is set at the ground reference point. W 、Y W and Z W The corresponding directions are along the robot's initial motion direction, extending perpendicular to the motion plane to the right, and vertically upward according to the right-hand rule. The quadruped robot's main coordinate system includes the center of mass coordinate system and the fuselage coordinate system. The center of mass coordinate system takes the robot's center of mass as its origin, its X-axis is always consistent with the instantaneous motion direction, its Y-axis points to the right side of the body according to the right-hand rule, and its Z-axis is vertically upward to form a right-hand coordinate system, which is mainly used for dynamic analysis and motion trend description. The fuselage coordinate system is fixed at the geometric center of the torso, its X-axis points to the front end along the longitudinal axis of the fuselage, its Y-axis points horizontally to the right, and its Z-axis points vertically upward according to the right-hand rule.

[0036] Ensuring initial axis alignment reduces the computational complexity of DH parameter transformation. Each leg of a quadruped robot consists of a hip joint, a thigh joint, and a shank joint. Servo motors drive the hip, thigh, and shank linkages, forming a complete kinematic chain. Each leg has a hip, thigh, and shank coordinate system. These three coordinate systems are fixed to the hip joint's rotation center, change orientation with thigh joint rotation, and are fixed to the shank linkage. Each of the four legs of the quadruped robot establishes a leg-base coordinate system along its extension direction. These leg-base coordinate systems maintain a fixed spatial relationship with the body coordinate system, with their origins symmetrically distributed within the body's transverse plane. Defining the coordinate systems of the hip and thigh joints at the same location, using a unified coordinate design, ensures initial axis alignment and reduces the computational complexity of DH parameter transformation. The origin of the shank coordinate system is set at the center of the knee joint and is initially aligned axially with the base coordinate system to ensure the continuity of motion parameter transmission.

[0037] S1-2. Based on the parameter characteristics of the mechanism, the DH parameter transformation method is used to construct the homogeneous transformation matrix between each joint, that is, to establish the mapping relationship between the joint angle and the foot end space.

[0038] S1-3. Based on each homogeneous transformation matrix, a spatial transformation model of adjacent joint coordinate systems is constructed.

[0039] S1-4. Use the Jacobian matrix to solve each spatial transformation model and obtain the spatial posture mapping relationship between the end of the leg actuator and the body base.

[0040] S1-5. Based on the spatial posture mapping relationship, perform mechanical and kinematic analysis on the quadruped robot, and use a single rigid body model to construct a quadruped robot mechanical motion model. The quadruped robot mechanical motion model includes a body mechanical motion model and a leg mechanical motion model.

[0041] The methods adopted by S1 are all existing technologies, so the corresponding details will not be elaborated in detail.

[0042] S2. Simulate the operation of a quadruped robot in complex terrain, and determine the quadruped robot's motion state estimation results, terrain estimation results, and foot-end contact estimation results based on the quadruped robot's mechanical motion model.

[0043] The S2 includes:

[0044] S2-1. Simulate the operation of a quadruped robot in a complex terrain and collect motion state parameters and joint torque sensor data. The motion state parameters include body motion information, angular motion information, and joint motion information.

[0045] The S2-1 includes:

[0046] S2-1-1. Use the inertial measurement unit to measure the original angular motion information of the quadruped robot in the body coordinate system in real time. The original angular motion information includes three-axis acceleration, angle data, and angular velocity data. The inertial measurement unit includes an accelerometer and a gyroscope. The accelerometer is used to provide the three-axis acceleration vector a in the body coordinate system. B The gyroscope is used to provide angle data and angular velocity data in the aircraft coordinate system. The angle data includes the angles of each joint corresponding to each leg (hip joint angle, thigh joint angle, and calf joint angle). The angular velocity data includes the angular velocity corresponding to each leg.

[0047] S2-1-2, using a state estimator to measure the quadruped robot's body motion information in the world coordinate system in real time. The body motion information includes the body position, body velocity, and the positions of the four legs.

[0048] S2-1-3. Measure the original joint motion information of each leg in the fuselage coordinate system in real time using the joint encoder. Each original joint motion information includes the joint positions (hip joint position, thigh joint position, and calf joint position) and joint velocities of the leg.

[0049] S2-1-4. Using the coordinate system conversion matrix in the quadruped robot's mechanical motion model, convert the original angular motion information and each original joint motion information in the body coordinate system into angular motion information and joint motion information in the world coordinate system. This conversion process is well known in the art and will not be described in detail.

[0050] S2-1-5. Collect joint torque data through sensors.

[0051] S2-2. Based on the motion state parameters, the motion state of the quadruped robot is estimated through the quadruped robot mechanical motion model and discrete Kalman filter to obtain the motion state estimation result.

[0052] The S2-2 includes:

[0053] S2-2-1. Based on the motion state parameters, construct the corresponding state transfer equation and discretize it to obtain the discrete state transfer equation.

[0054] S2-2-2. Based on the motion state parameters and the quadruped robot's mechanical motion model, construct the observation equation for the quadruped robot. The discrete state transfer equation and observation equation are both existing technologies, so they will not be described in detail.

[0055] S2-2-3. Use the discrete Kalman filter to perform state estimation on the discrete state transfer equation and observation equation to obtain the motion state estimation result. The corresponding formula is:

[0056]

[0057] Among them, A k represents the state transfer matrix, B k represents the control matrix, u k-1 Represents the observation vector input value (three-axis acceleration) at time k-1, Represent the discretized optimal state vectors (Value of the discrete state transition equation) Prior estimate at time k, posterior estimate at time k+1, posterior estimate at time k, denote the prior covariance and posterior covariance at time k, respectively. represents the posterior covariance at time k-1, Q represents the covariance matrix of Gaussian noise, (·) T represents the transposed matrix, R represents the noise matrix, C represents the observation matrix, and K represents the transformation matrix of the quadruped robot mechanical motion model. k represents the Kalman filter gain, y k represents the value of the observation equation at time k, and I represents the identity matrix.

[0058] S2-3. Construct a terrain plane model based on the motion state parameters; calculate ground irregularity through the terrain plane model and use it as a terrain estimation result.

[0059] The S2-3 includes:

[0060] S2-3-1. Select the motion state parameters and choose the four foot positions during the support phase. Determine the coefficient matrix of the original plane equation of the quadruped robot, such as Figure 2 As shown. Among them, p Wi Indicates the four foot positions in the world coordinate system. When i takes the values of 0, 1, 2 and 3 respectively, The corresponding positions of the feet of the left front leg, right front leg, right hind leg and left hind leg.

[0061] S2-3-2. Based on the positions of each foot end and the coefficient matrix of the original plane equation, a terrain plane model is constructed. The corresponding formula is:

[0062]

[0063] E = [e0, e1, e2];

[0064] in, represents the height coordinate of the i-th foot end of the quadruped robot in the support phase, represents the coordinate of the i-th foot end of the quadruped robot on the horizontal plane in the world coordinate system when it is in the support phase, E represents the coefficient matrix of the original plane equation, and e0, e1, and e2 represent the coefficients of the coefficient matrix of the original plane equation.

[0065] When the i-th foot of the quadruped robot is in the support phase, the foot coordinate uses the current position; when the i-th foot of the quadruped robot is in the swing phase, the foot position of the previous cycle is selected.

[0066] S2-3-3. Solve the terrain plane model using the least squares method to obtain the initial plane equation coefficient matrix E1. The formula corresponding to the initial plane equation coefficient matrix E1 is:

[0067]

[0068] S2-3-4. Use the sliding filter based on discrete time series to filter the initial plane equation coefficient matrix E1 to obtain the plane equation coefficient matrix The corresponding formula is:

[0069]

[0070] Among them, ∑(·) represents the summation function, α j represents the weight coefficient at time j, n represents the time length, E k-i Represents the initial plane equation coefficient matrix at time kj. Weight coefficient α j It is adjusted in a linear increasing manner, that is, α j =nj.

[0071] S2-3-5. Based on the plane equation coefficient matrix Calculate the degree of terrain irregularity, which is the terrain estimation result. The formula corresponding to the degree of terrain irregularity is:

[0072]

[0073]

[0074] in, Represents the plane equation coefficient matrix The coefficient in δ i R represents the residual difference between the actual measured height of the i-th foot end and the predicted height of the terrain plane, δ Represents the degree of terrain irregularity, and N represents the total number of contact points when a single leg is used as the supporting leg. i It can indicate the local concavity and convexity of the foot contact point. The larger the absolute value, the more significant the terrain undulation at that point.

[0075] S2-4. Construct a robot foot contact probability model and solve it to obtain the foot contact estimation result.

[0076] The S2-4 includes:

[0077] S2-4-1. Define the normalized phase variable. During the movement of the quadruped robot, the foot end presents the characteristics of alternating ground support and air swinging. Based on the analysis of the dynamic characteristics of the gait phase, the original normalized phase variable φ is defined. The introduction of φ can effectively eliminate the influence of the gait cycle length difference on the state estimation, and at the same time provide a unified timing benchmark for multi-legged coordinated movement. By introducing the phase offset φ i,offset , which can realize multi-leg coordinated control, and each leg has its own normalized phase variable φ i , thus, φ i The corresponding formula is:

[0078] φ=(t-t0) / T;

[0079] φ i =φ i,offset +φ;

[0080] Where t and t0 represent the current time and the initial time, respectively, and T represents a gait cycle. The gait cycle is the gait phase period, which is the time required for a quadruped robot to complete a complete motion cycle including the stance phase and the swing phase on a single leg.

[0081] S2-4-2. Based on the normalized phase variable, the mapping relationship between the leg motion phase and the contact probability of the quadruped robot is determined through the Gaussian mixture probability distribution function, and the first foot-end contact probability model is constructed, that is, a contact probability model based on the gait phase is constructed.

[0082] During the operation of a quadruped robot, the legs act as a supporting structure in the support phase to bear the load of the body, and the contact probability between the foot and the ground is close to 1, providing support and motion stability for the body. The swing phase plans the motion trajectory of the foot to make it move in the air, and the contact probability between the foot and the ground is close to 0. In actual motion, affected by the robot's own dynamic characteristics and the terrain, such as foot slippage or slight deformation of the ground, in the transition area between the support phase and the swing phase, the contact state of the foot will show the characteristics of a probability transition in the critical region of phase change. By constructing a Gaussian mixture probability distribution function to describe the mapping relationship between the leg motion phase and the contact probability, the corresponding formula of the contact probability model based on the gait phase is:

[0083]

[0084] Where c is a constant, φ c0 、φ c1 denote the gait phases corresponding to the stance phase c0 and the swing phase c1, erf(·) denotes the error function, μ c0 , σ c0 represent the mean and variance of the support phase, μ c1 , σ c1 denote the mean and variance of the swing phase, P(c|φ c0 )、P(c|φ c1 ) represent the foot contact probability corresponding to the stance phase c0 and the swing phase c1, represents the phase state corresponding to the normalized phase variable of the ith foot end, Represents the first foot end contact probability based on the gait phase (the output of the first foot end contact probability model).

[0085] like Figure 3 As shown, it shows that the quadruped robot is in the support phase time interval, φ i When approaching the midpoint of the support phase period, It shows deterministic contact characteristics; the critical region of phase change at the beginning and end of the support phase indicates that there is uncertainty in the contact state during this period. In the swing phase time interval, when φ i When approaching the midpoint of the swing phase cycle, the foot end movement reaches the maximum lifting height. Converge to the off-ground state.

[0086] S2-4-3. Based on the joint torque data, a second foot-end contact probability model is constructed through the Jacobian matrix and the Gaussian probability model, that is, a foot-end contact probability model based on the foot-end force is constructed.

[0087] Using joint torque data and based on the physical relationship between force and contact, the foot-end contact probability based on the foot-end force is established. Under different terrain conditions, the performance of the foot-end contact probability model based on the gait phase varies. On flat ground, the model can more accurately predict the contact state of the foot, and the calculated value of the contact probability is highly consistent with the actual situation. When the robot moves on a rough road, due to the irregular undulations of the ground, the foot-end touches the ground early or late, and the foot-end force is closely related to the foot-end contact. Calculate the foot-end force f z , using the direct relationship between foot end velocity and joint angular velocity, the Jacobian matrix between the two can realize the mutual solution of joint torque and foot end force, and build a contact force measurement model. z Create a corresponding Gaussian probability model based on the contact force measurement model, define the contact probability and the ground height model measurement value to form an observation vector, and then the formula corresponding to the foot-end contact probability model based on the foot-end force is:

[0088]

[0089] Among them, P(c|f z,i ) represents the second foot end contact probability based on the foot end force (output of the second foot end contact probability), μf z,i ,σf z,i They represent the current expected foot end force and foot end force standard deviation of the i-th foot end, respectively, z,i Indicates the current foot force of the i-th foot in the Z direction in the world coordinate system.

[0090] like Figure 4 As shown in the figure, under different variances, the corresponding probability curve presents an "S" shape, with the foot end force f z,i The increase of P(c|f z,i ) gradually increases. When f z,i After reaching a certain value, P(c|f z,i ) approaches 1.

[0091] S2-4-4. Obtain the height corresponding to the projection position of the foot end position on the terrain plane based on the terrain plane model Construct the third foot contact probability model, that is, construct the probability model of foot height and foot contact with the ground. As the foot end height of the quadruped robot in the Gaussian distribution during the movement process obeys the mean, the variance of the Z direction coordinate of the quadruped robot in the world coordinate system is As the variance of the Gaussian distribution. Therefore, the formula corresponding to the probability model of foot end height and foot end touching the ground is:

[0092]

[0093] Among them, p Wi,z represents the Z-direction coordinate of the i-th foot end in the world coordinate system, that is, the foot end height, P(c|p Wi,z ) represents the third foot contact probability of the foot end height and the foot end touching the ground (the output of the third foot end contact probability model).

[0094] like Figure 5 As shown in the figure, under different variances, if the quadruped robot walks on flat ground, Always zero, p Wi,z The closer to the ground, the higher the P(c|p Wi,z ) is higher.

[0095] S2-4-5. Use the Kalman filter fusion multi-probability model method to fuse the first foot-end contact probability model, the second foot-end contact probability model and the third foot-end contact probability model to construct a foot-end contact probability model.

[0096] The S2-4-5 includes:

[0097] Based on the first foot end contact probability model, a complete observation space for state estimation is established, that is, a Kalman filter model for quadruped robot state estimation is established. The corresponding formula is:

[0098]

[0099] Among them, u k Represents the observation vector input value at time k, x k 、x k-1 Represents the state matrix at time k and time k-1, y′ k represents the expected value vector, C k Represents the identity matrix. Where A k =04, A k is a fourth-order zero matrix; B k =I4,B k is the fourth-order identity matrix, They represent the contact probabilities based on the gait phase corresponding to the left foreleg, right foreleg and the i-th foot of the quadruped robot respectively.

[0100] The second foot end contact probability model and the third foot end contact probability model are used as measurement values, and the second contact probability model and the third contact probability model are fused to obtain the fused contact probability model P mix , the corresponding formula is:

[0101] P mix =[(1-β)P(c|f z,i )+βP(c|p Wi,z )];

[0102]

[0103] Where β represents the measurement weight. The gaits of the quadruped robot are divided into diagonal trotting, slow walking, ipsilateral trotting and crawling gaits.

[0104] The fused contact probability model and Kalman filter model are processed using a low-pass filter to obtain the foot-end contact probability model. The formula corresponding to the foot-end contact probability model is:

[0105] y′ k,i =0.8·P mix,k,i +(1-0.8)·P mix,k-1,i ;

[0106] Among them, y′ k,i represents the foot contact probability of the i-th foot at time k (the output of the foot contact probability model), P mix,k,i represents the output of the fused contact probability model of the i-th foot at time k, P mix,k-1,i Represents the output of the fused contact probability model for the i-th foot at time k-1.

[0107] The present invention integrates multi-source data from an inertial measurement unit and a joint encoder, adopts a Kalman filter algorithm to estimate the body posture and velocity, and effectively suppresses the interference of sensor noise on the motion parameter calculation; establishes a terrain plane model based on the geometric distribution of the supporting foot end, calculates the terrain inclination characteristics in real time, and provides a terrain parameter benchmark for gait planning; proposes a multimodal probability model that integrates gait phase, joint torque, and foot end height, and estimates the foot-ground contact state through a weighted strategy, which provides theoretical support for improving the motion adaptability of quadruped robots in unstructured terrain.

[0108] S3. Based on the motion state estimation results, terrain estimation results and foot contact estimation results, the gait of the quadruped robot is adjusted in real time.

[0109] When a quadruped robot moves, the foot end is divided into a support phase SP1 in contact with the ground and a swing phase SP2 in the air. In the support phase (i.e., the foot end is grounded), the legs bear the weight and balance, while in the swing phase, the legs are responsible for trajectory planning and position adjustment. When the quadruped robot switches gaits, if it switches suddenly during the swing phase, the swing leg will suddenly fall to the ground or the support leg will suddenly lift up, causing an impact on the body and affecting the stability of the movement. In order to achieve a smooth transition, the current cycle gait should be completed. After the swing leg stably touches the ground, the support leg of the previous cycle is used as the reference leg for the gait after the transition, and then the swing phase is started. Since the parameters of each gait phase are independent, it is difficult to achieve instantaneous switching by directly changing the parameters. Therefore, similar gaits are set as adjacent motion states, that is, diagonal walking is used as an adjacent motion state, and the motion stability is improved by switching between adjacent gaits.

[0110] Thus, the S3 includes:

[0111] S3-1. Perform trajectory planning on the quadruped robot to obtain an operation planning scheme. The operation planning scheme includes a gait phase planning scheme, a foot landing point planning scheme, and a foot end trajectory planning scheme.

[0112] The core of quadruped robot gait planning is to coordinate the leg movement timing, plan the landing point and design a smooth foot trajectory. The present invention plans from three aspects: gait phase, landing point and foot trajectory. A periodic gait is constructed by duty cycle and phase offset, and the leg movement is decomposed into alternating support and swing phases to form standard gaits such as diagonal trotting; the landing point is adjusted in real time by combining speed interpolation and posture estimation to adapt to translation and rotation scenarios; a segmented trajectory is designed for the swinging leg, and cubic curve fitting is used for the lift-off, take-off and touchdown stages, and delay compensation is introduced to correct the trajectory to avoid foot-end stepping and jamming. Thus, the S3-1 includes:

[0113] S3-1-1. Perform gait phase planning on the quadruped robot to obtain a gait phase planning scheme.

[0114] When a quadruped robot moves, the four legs alternately switch between two states in a fixed sequence, and different phase combinations form a periodic gait pattern. The gait cycle is the support phase period T st and the swing phase period T sw The duty cycle is the proportion of the stance phase period in the gait cycle.

[0115] In the present invention, the duty cycle γ of the quadruped robot is calculated; based on the normalized phase variable φ i , calculate the phase progress φ of the support phase and the swing phase s ; Based on γ and φ s , set the gait phase planning scheme of the quadruped robot, as shown in Table 1, the corresponding gait phase is as follows Figure 6 As shown. Figure 6 In the figure, orange indicates that the foot is in the swing phase, and white indicates that the foot is in the stance phase.

[0116] Table 1 Gait phase planning scheme

[0117] gait Duty cycle Phase offset diagonal trot [0.50，0.50，0.50，0.50] [0.00，0.50，0.50，0.00] slow walk [0.75，0.75，0.75，0.75] [0.00，0.25，0.50，0.75] Same-side trot [0.50，0.50，0.50，0.50] [0.00，0.50，0.00，0.50] Walking diagonally [0.70，0.70，0.70，0.70] [0.00，0.50，0.50，0.00]

[0118] S3-1-2. Plan the landing point of the quadruped robot and obtain a landing point planning scheme.

[0119] The quadruped robot's single leg is considered an inverted pendulum model. When the connecting rod is vertical, the line of gravity passes through the support point, and the inverted pendulum model is stable. The center of mass of the fuselage is roughly located at the origin of the fuselage coordinate system, and the symmetrical landing point is set as the horizontal coordinate of the thigh joint in the fuselage coordinate system. The robot leg adopts a full elbow structure, which causes the center of mass to shift during movement, so the symmetrical landing point correction value p is introduced. offset , correct the symmetrical foot point, the corresponding formula is: p b =p hip +p offset ; Among them, p b 、p hip They represent the corrected symmetrical landing point and the symmetrical landing point respectively.

[0120] Assume that the quadruped robot moves at a constant speed, and calculate the next symmetrical landing point according to the corrected symmetrical landing point The corresponding formula is: Among them, Δx1 and Δx2 represent the moving distances between any two of the three adjacent motion states, and pb(0) represents the corrected symmetrical landing point in the support phase.

[0121] based on Calculate the real-time coordinates of the quadruped robot at the current moment in real time. The corresponding formula is:

[0122]

[0123] in, represents the coefficients of the plane equation coefficient matrix, Indicates the coordinates of the foot landing point of the quadruped robot at the current moment (in the world coordinate system), xb(SP1), y b (SP1) represents the coordinate of the corrected symmetrical foothold in the horizontal plane under the support phase SP1, V x 、 Represents the x component of the quadruped robot's speed and its corresponding mapping component, Represents the velocity interpolation coefficient, V y 、 They represent the y component of the quadruped robot’s speed and its corresponding mapping component, represents the angle between the foot ends between adjacent motion states, cos(·) and sin(·) represent the cosine function and sine function respectively, and R1 represents the projection length from the origin of the fuselage coordinate system to the symmetrical foot landing point in the horizontal and vertical planes.

[0124] Repeatedly update the next symmetrical landing point and real-time coordinates until the complex terrain simulation is completed, and obtain the landing point planning scheme.

[0125] S3-1-3. Perform foot-end trajectory planning on the quadruped robot to obtain a foot-end trajectory planning solution.

[0126] The foot trajectory of a quadruped robot's swinging leg is always the trajectory from the support point of the previous cycle to the landing point of this cycle. There are many variations in the calculation method of the foot trajectory, but its trajectory during the flight process must play a positive role in the robot's movement. The foot trajectory is divided into three stages. The first and second stages are the normal swing cycle, both of which include the lift-off stage and the landing stage. The third stage is the exploration landing point trajectory planned after the delayed touchdown. When the foot begins to lift off, the foot must be kept at the X coordinate system in the world coordinate system. W Y W The plane displacement is uniform and the desired foot swing height is reached quickly; if the foot contact estimate indicates a delayed contact, the descent phase is planned from that point.

[0127] Thus, S3-1-3 includes:

[0128] Based on the energy optimization principle, the swing trajectory function is constructed, and the corresponding formula is:

[0129] x track =a1t 3 +b1t 2 +c1t+d1;

[0130]

[0131] Among them, a1, b1, c1, d1 represent coefficients, t represents time, x track 、 represent the swing trajectory displacement and its first-order derivative respectively.

[0132] Set trajectory boundary conditions, including a first trajectory boundary condition in the lift-off phase and the descent phase, a second trajectory boundary condition, and a third trajectory boundary condition for triggering delayed touchdown.

[0133] The boundary condition of the first trajectory is: X in the world coordinate system W Y W For a plane, when the quadruped robot swings at a unit step length, the displacement of the foot end at the starting point of the lifting phase (when it lifts off the ground) is 0, and the displacement of the end point of the falling phase (when it touches the ground) is 1;

[0134] The boundary condition of the second trajectory is: Z in the world coordinate system W Direction: When the quadruped robot avoids obstacles during the ascent phase, the velocity direction of its foot changes, and the velocity at the highest point is also 0;

[0135] The third trajectory boundary condition is: after the quadruped robot triggers delayed touchdown, Z W The constraints of the foot end position and speed are calculated based on the preset delayed swing phase, and its XW Y W The direction only makes a small displacement forward, and its displacement can be regarded as 0.

[0136] The swing trajectory function is optimized based on the trajectory boundary conditions to obtain the swing trajectory optimization function. The corresponding formula is:

[0137]

[0138] in, Represents the optimized swing trajectory.

[0139] The swing trajectory optimization function is used to calculate the swing trajectory of the quadruped robot under different gaits in the process of simulating the quadruped robot running in complex terrain, and the foot end trajectory planning scheme is obtained.

[0140] S3-2. Set the sampling frequency; start the quadruped robot and implement the operation planning scheme; based on the sampling frequency, use the same method as S2-3 to calculate the average ground irregularity in the gait cycle corresponding to the current gait in real time, and collect the terrain inclination angle of the current gait in real time. In this embodiment, the sampling frequency is 500 times / second. During the gait cycle, the corresponding ground irregularity is calculated in real time according to the sampling frequency, and the average of multiple ground irregularities is calculated. Since the ground inclination angle changes steadily in the current gait, the next terrain inclination angle of the current gait can be collected. If higher accuracy is required, the terrain inclination angle of the gait cycle under the current gait can be collected and the corresponding terrain inclination angle average can be calculated.

[0141] S3-3, based on the terrain-gait mapping table, the current gait and the corresponding ground irregularity average and the fixed terrain inclination angle, obtain the terrain type under the current gait. The terrain-gait mapping table is shown in Table 2. In Table 2, |θ pitch | represents the fixed terrain inclination angle θ pitch The absolute value of R δ * Represents the mean value of ground irregularity.

[0142] Table 2 Terrain-gait mapping table

[0143]

[0144] S3-4. Calculate the motion stability of the quadruped robot in the current gait. The formula corresponding to the motion stability DSI is:

[0145] DSI = α1ΔP + α2Δθ;

[0146] Where α1 and α2 represent weight coefficients, which are 0.2 and 0.8 respectively. ΔP represents the average value of the fuselage position from the expected fuselage position within a gait cycle. Δθ represents the average value of the fuselage pitch angle from the expected fuselage pitch angle within a gait cycle.

[0147] S3-5. Based on the terrain type and movement stability of the current gait, select the next gait as the candidate gait.

[0148] When a quadruped robot switches gaits, if it suddenly switches during the swing phase, the swing leg will suddenly land or the supporting leg will suddenly lift up, causing an impact on the body and affecting the stability of the movement. To achieve a smooth transition, the current cycle gait should be completed. After the swing leg has stably touched the ground, the supporting leg of the previous cycle is used as the reference leg for the switched gait, and then the swing phase is started. Because the phase parameters of each gait are independent, it is difficult to achieve instantaneous switching by directly changing the parameters. Therefore, similar gaits are set as adjacent motion states, and the motion stability is improved by switching between adjacent gaits. In this embodiment, the similar gait is diagonal walking.

[0149] The relationship between stability and gait is shown in Table 3.

[0150] Table 3 Stability-gait mapping table

[0151] Terrain-Gait Walking diagonally diagonal trot slow walk Same-side trot Smooth surface DSI<0.10 DSI<0.10 DSI<0.10 DSI<0.10 Slippery slope DSI<0.30 DSI<0.50 DSI<0.30 ∞ stairs DSI<0.50 ∞ DSI<0.40 ∞ Concave and convex surfaces DSI<0.15 DSI<0.10 DSI<0.15 DSI<0.15

[0152] Based on the stability-gait mapping table, the terrain type under the current gait, and the motion stability, the next gait is selected as the candidate gait. Because the robot's maximum speed is closely related to its gait planning, it automatically switches to a higher-frequency, faster ipsilateral trot on flat ground. When traversing uneven surfaces, the robot chooses a diagonal trot, which also offers higher speed and better stability. When sensing a slope, the robot first switches to a more stable diagonal walking gait, then selects a corresponding gait based on the terrain it senses after three gait cycles.

[0153] S3-6. Based on the foot end contact estimation results and the motion state estimation results, the selected gait is optimized to obtain a gait optimization strategy.

[0154] S3-6 includes:

[0155] S3-6-1. Determine the contact state of the current gait based on the foot contact estimation results and the motion state estimation results. Contact states include the support phase, delayed contact, advanced contact, and swing phase. Determining the contact state of a quadruped robot is the core foundation of gait optimization, and its accuracy directly affects its terrain adaptability. The contact state of the quadruped robot's foot is determined using the fused contact probability. Thus, S3-6-1 includes:

[0156] S3-6-1-1. Determine the ground contact threshold through simulation experiments; determine the contact probability of the current gait through the foot-end contact estimation results; and calculate the foot-end velocity of the current gait through the motion state estimation results.

[0157] S3-6-1-2, determine whether the contact probability is less than the touchdown threshold; if so, determine the first contact threshold s p is 0; otherwise, the first contact threshold s is determined p is 1.

[0158] S3-6-1-3, determine whether the foot end speed is 0; if so, determine the second contact threshold s φ is 1; otherwise, the second contact threshold s is determined φ is 0.

[0159] S3-6-1-4, determine whether the foot end speed is greater than the speed tolerance threshold; if so, determine the third contact threshold s v is 0; otherwise, the third contact threshold s is determined v is 1.

[0160] S3-6-1-5, based on the first contact threshold, the second contact threshold and the third contact threshold, determine the contact state of the current gait. p =1s φ =1 or s p =1,s φ ≠1,s v =1, the contact state is the support phase; when s p =1,s φ ≠1,s v ≠1, the contact state is delayed contact; when s p ≠1,s φ ≠1 or s p ≠1,s φ =1,s v ≠1, the contact state is swing phase; when s p ≠1,s φ =1,s v =1, the contact state is early contact.

[0161] S3-6-2. Based on the contact state, the gait to be selected is optimized, that is, the stance phase period and the swing phase period in the gait to be selected are adjusted to obtain a gait optimization strategy.

[0162] During the dynamic motion of a quadruped robot, timing deviations in foot contact events can lead to phase misalignment between the support and swing phases, potentially causing stability issues such as center of mass trajectory deviation and sudden changes in joint torque. Traditional gait generation methods based on fixed phase allocations exhibit limitations in such scenarios. Their rigid timing constraints cannot adapt to the uncertainty in touchdown timing caused by terrain irregularities. The gait cycle of a quadruped robot consists of alternating support and swing phases, and its core lies in the precise control of the phase difference and timing of each leg. On ideal flat terrain, the robot can achieve stable periodic motion using pre-set gait parameters such as cycle duration, duty cycle, and phase deviation. However, in irregular terrain, sudden changes in ground height or friction characteristics can cause premature or delayed foot contact. If fixed gait parameters are used, premature contact will cause the swing leg to prematurely terminate trajectory tracking and enter the support state, resulting in sudden changes in center of mass acceleration. Delayed contact will prevent the swing leg from providing support force in a timely manner, leading to the risk of the robot tipping over due to prolonged suspension of one leg. Therefore, S3-6-2 includes:

[0163] By adjusting the duration of the support phase and swing phase in the selected gait in real time, the global consistency of the gait cycle of the selected gait and the stability of the coordinated movement of each leg are ensured, and the following conditions are met: (1) When the single-leg touchdown time is abnormal, the global gait cycle is maintained unchanged through local phase adjustment to avoid multi-leg movement synchronization; (2) After the touchdown state suddenly changes, the support and swing phase distribution is quickly reconstructed to ensure the continuity of the robot's main body movement; (3) The phase compensation mechanism eliminates the impact of a single adjustment on the subsequent gait cycle and ensures long-term movement stability.

[0164] When the contact state of the i-th leg is delayed touchdown, the leg needs to extend the execution time of the swing phase trajectory in the selected gait, and correspondingly extend the corresponding support phase to maintain the overall cycle unchanged; when the contact state of the i-th leg is early touchdown, it is necessary to immediately terminate the current swing phase and enter the support phase, while extending the next support cycle to ensure cycle stability. The optimized foot trajectory is effective in dealing with early touchdown and delayed touchdown situations. Figure 7 shown.

[0165] In the experiment, the gait period of the quadruped robot is T, the duty cycle is γ, and the phase offset is φ i,offset and the current time is t k , the theoretical end time of the swing phase of the i-th leg is When the contact state of the i-th leg is delayed contact, the swing phase time Δt of the leg in the selected gait is extended. i =αT, making the current cycle swing phase period T sw =(1-γ)T+Δt i , while compressing the stance phase period T in the candidate gait st =γT-Δti , in order to maintain the overall gait cycle T unchanged; when the contact state of the i-th leg is early touchdown, the current swing phase is terminated immediately, and its unfinished swing time The support period T of the leg in the selected gait st =γT+Δt i Among them, the extension time coefficient α satisfies: They represent the start time and end time of the theoretical swing phase of the i-th leg respectively.

[0166] S3-7. Based on the gait optimization strategy, the gait of the quadruped robot is controlled.

[0167] The present invention constructs a three-dimensional terrain plane equation through real-time terrain perception and kinematic modeling, combines speed instructions with state observation data to predict the center of mass motion trajectory in the future time domain, and establishes a terrain-adaptive posture adjustment mechanism. A single rigid body dynamics model is used to construct a linearized state space equation, and the complex dynamic constraints are converted into matrix operations that can be calculated in real time through discretization processing. The friction cone constraint and gait phase coordination mechanism are introduced to transform the foot-end force distribution problem into a constrained quadratic programming problem. Through a closed-loop architecture of rolling time domain optimization and feedback correction, the optimal control sequence of the foot-end force is solved. Since S3-7 are existing technologies, they are briefly described without further elaboration.

[0168] like Figure 8 As shown, a quadruped robot gait control system based on complex terrain includes:

[0169] The quadruped robot mechanical motion model construction module is used to obtain the mechanism parameter characteristics of the quadruped robot and construct the quadruped robot mechanical motion model.

[0170] The quadruped robot mechanical motion model simulation module is used to simulate the operation of the quadruped robot in complex terrain, and based on the quadruped robot mechanical motion model, determine the quadruped robot's motion state estimation results, terrain estimation results and foot end contact estimation results.

[0171] The real-time gait adjustment module of the quadruped robot is used to adjust the gait of the quadruped robot in real time based on the motion state estimation results, terrain estimation results and foot end contact estimation results.

[0172] In this embodiment, the method is used to control the quadruped robot, and the quadruped robot is controlled to walk slowly at a constant speed along the X coordinate system of the world. W The robot walks at a speed of 0.2 m / s and collects the curves of the expected position of the robot's center of mass and the estimated state position, such as Figure 9 and Figure 10As shown in the figure, during the robot's straight-line walking at a speed of 0.2 m / s along the x-axis, the center of mass trajectory estimated through motion state estimation tracks the desired trajectory well, with the deviation between the two maintained at a low level. Actual measurement data shows that the position offset along the x-axis is always within the engineering error. The periodic fluctuations exhibited by the velocity estimation curve are significantly correlated with the foot contact impact during the gait cycle, and its oscillation amplitude remains stable within 0.1 m / s of the set speed. This demonstrates the reliability of the quadruped robot's motion state estimation in calculating the robot's current position and velocity.

[0173] Taking a steep slope as an example, control the quadruped robot to walk on the steep slope and record the relevant variables calculated by the terrain plane equation. Figure 11 and Figure 12 As shown in the figure, when a quadruped robot transitions from flat ground to an inclined slope, the pitch angle of the terrain changes after the foot of the swing phase leg lands, causing the data used in the plane equation to change. This shifts the spatial coordinates of the foot of the supporting phase, prompting the terrain estimation module to respond in real time. Once the robot has fully landed on the slope, its estimated value converges to the actual pitch angle, verifying the dynamic response and accuracy of the terrain plane equation during the quadruped robot's motion.

[0174] On a smooth surface, the quadruped robot moves at a speed of 0.2 m / s and in slow walking and diagonal walking respectively, and the driving time is 10 s. The RMSE index of the quadruped robot in two gaits (slow walking and diagonal walking) is recorded, as shown in Table 4.

[0175] Table 4 RMSE index table

[0176]

[0177]

[0178] As shown in Table 4, the RMSE values for both gaits are less than 0.1, indicating that the model predictive control can effectively calculate the plantar force of the robot during the support phase on uneven terrain, realize the tracking of the predicted trajectory during the robot's motion, and provide a reliable control framework for the optimization of subsequent motion control strategies.

[0179] In summary, the present invention is based on the parameter characteristics of the quadruped robot mechanism, establishes its dynamic and kinematic mathematical models and establishes a quantitative method for the degree of terrain unevenness, so as to realize the estimation of the contact state probability of the foot end during the robot movement; for the robot movement under irregular terrain, different gait types are designed to match the irregular terrain, and the stability and switching strategy of the gait under different terrains are analyzed; based on the contact state probability of the foot end, the contact state of the foot end is judged and the real-time optimization of the gait phase is realized, so as to realize the dynamic balance control of the quadruped robot during the movement.

[0180] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A gait control method for a quadruped robot based on complex terrain, characterized in that: include: Obtain the mechanism parameter characteristics of the quadruped robot and construct the mechanical motion model of the quadruped robot; Simulate the operation of a quadruped robot in complex terrain, and determine the quadruped robot's motion state estimation results, terrain estimation results, and foot end contact estimation results based on the quadruped robot's mechanical motion model; Based on the motion state estimation results, terrain estimation results and foot contact estimation results, the gait of the quadruped robot is adjusted in real time.

2. The gait control method of a quadruped robot based on complex terrain according to claim 1, characterized in that: The method of obtaining the mechanical parameter characteristics of the quadruped robot and constructing the mechanical motion model of the quadruped robot includes: Acquiring mechanism parameter characteristics and setting multiple coordinate systems; the mechanism parameter characteristics include link translation, joint angle, link rotation offset and link extension; Based on the parameter characteristics of the mechanism, the DH parameter transformation method is used to construct the homogeneous transformation matrix between each joint, that is, to establish the mapping relationship between the joint angle and the foot end space; Based on each homogeneous transformation matrix, a spatial transformation model of adjacent joint coordinate systems is constructed; The Jacobian matrix is used to solve the spatial transformation models to obtain the spatial pose mapping relationship between the leg actuator end and the body base; Based on the spatial posture mapping relationship, the mechanical and kinematic analyses of the quadruped robot are carried out, and the mechanical motion model of the quadruped robot is constructed using a single rigid body model.

3. The gait control method of a quadruped robot based on complex terrain according to claim 1, characterized in that: The simulating operation of the quadruped robot in the complex terrain and determining the motion state estimation result, terrain estimation result and foot end contact estimation result of the quadruped robot based on the mechanical motion model of the quadruped robot include: Simulate the operation of a quadruped robot in complex terrain and collect motion state parameters and joint torque sensing data; Based on the motion state parameters, the motion state of the quadruped robot is estimated through the quadruped robot mechanical motion model and discrete Kalman filter to obtain the motion state estimation result; Construct a terrain plane model based on the motion state parameters; calculate the ground irregularity through the terrain plane model and use it as the terrain estimation result; A robot foot contact probability model is constructed and solved to obtain the foot contact estimation result.

4. The gait control method of a quadruped robot based on complex terrain according to claim 3, characterized in that: The terrain plane model is constructed based on the motion state parameters; the ground irregularity is calculated by the terrain plane model and used as the terrain estimation result; The motion state parameters are selected, and the four foot end positions during the stance phase are selected; Determine the coefficient matrix of the original plane equation of the quadruped robot; Construct a terrain plane model based on the positions of each foot end and the coefficient matrix of the original plane equation; Solve the terrain plane model by the least square method to obtain the coefficient matrix of the initial plane equation; The initial plane equation coefficient matrix is filtered by using a sliding filter based on discrete time series to obtain the plane equation coefficient matrix; Based on the plane equation coefficient matrix, the degree of terrain irregularity is calculated, that is, the terrain estimation result.

5. The gait control method of a quadruped robot based on complex terrain according to claim 3, characterized in that: The process of constructing the robot foot contact probability model includes: Define the normalized phase variable; Based on the normalized phase variable, the mapping relationship between the leg motion phase and the contact probability of the quadruped robot is determined by the Gaussian mixture probability distribution function, and the first foot end contact probability model is constructed; Based on the joint torque data, the second foot contact probability model is constructed using the Jacobian matrix and Gaussian probability model; Based on the terrain plane model, the height corresponding to the projection position of the foot end position on the terrain plane is obtained to construct a third foot end contact probability model; The Kalman filter fusion multi-probability model method is used to fuse the first foot end contact probability model, the second foot end contact probability model and the third foot end contact probability model to obtain the foot end contact probability model.

6. The gait control method of a quadruped robot based on complex terrain according to claim 1, characterized in that: The method of adjusting the gait of the quadruped robot in real time based on the motion state estimation result, the terrain estimation result, and the foot end contact estimation result includes: Performing trajectory planning on the quadruped robot to obtain an operation planning scheme; the operation planning scheme includes a gait phase planning scheme, a foot landing point planning scheme, and a foot end trajectory planning scheme; Set the sampling frequency; start the quadruped robot and implement the operation plan; based on the sampling frequency, calculate the ground irregularity of the current gait in real time and collect the terrain inclination angle of the current gait in real time; Obtain the terrain type under the current gait based on the terrain-gait mapping table, the current gait, the corresponding ground irregularity average value and the fixed terrain inclination angle; Calculate the motion stability of the quadruped robot in the current gait; select the next gait as a candidate gait based on the terrain type and motion stability under the current gait; Based on the foot contact estimation results and motion state estimation results, the selected gait is optimized to obtain the gait optimization strategy; The gait of the quadruped robot is controlled based on the gait optimization strategy.

7. The gait control method of a quadruped robot based on complex terrain according to claim 6, characterized in that: The process of obtaining the foot end trajectory planning scheme includes: Based on the energy optimization principle, the swing trajectory function is constructed; Setting trajectory boundary conditions; the trajectory boundary conditions include a first trajectory boundary condition in the lift-off phase and a second trajectory boundary condition in the descent phase, and a third trajectory boundary condition for triggering delayed touchdown; Optimizing the swing trajectory function based on the trajectory boundary conditions to obtain the swing trajectory optimization function; The swing trajectory of the quadruped robot under different gaits is calculated through the swing trajectory optimization function to obtain the foot end trajectory planning scheme.

8. The gait control method of a quadruped robot based on complex terrain according to claim 6, characterized in that: The step of optimizing the selected gait based on the foot end contact estimation result and the motion state estimation result to obtain a gait optimization strategy includes: Determine the contact state of the current gait based on the foot end contact estimation result and the motion state estimation result; Determine the extension time coefficient; based on the contact state and the extension time coefficient, adjust the gait period of the selected gait, extend or shorten the swing phase period and the stance phase period of the selected gait, and complete the optimization of the selected gait.

9. A quadruped robot gait control system based on complex terrain, used to implement a quadruped robot gait control method based on complex terrain as claimed in any one of claims 1 to 8, characterized in that: include: A quadruped robot mechanical motion model construction module is used to obtain the mechanism parameter characteristics of the quadruped robot and construct the quadruped robot mechanical motion model; The quadruped robot mechanical motion model simulation module is used to simulate the operation of the quadruped robot in complex terrain and determine the quadruped robot's motion state estimation results, terrain estimation results, and foot end contact estimation results based on the quadruped robot's mechanical motion model; The real-time gait adjustment module of the quadruped robot is used to adjust the gait of the quadruped robot in real time based on the motion state estimation results, terrain estimation results and foot end contact estimation results.

Images