A self-adaptive gait planning and posture control method for a hexapod robot based on asymmetric narrow space constraints

Abstract

The application discloses a kind of based on asymmetric narrow space constraint hexapod robot adaptive gait planning and posture control method, belong to the motion control technical field of multi-legged robot.Step 1: obtain environment information and body posture, carry out asymmetric support surface modeling;Step 2: hexapod robot controller based on the modeling of step 1 and the roll angle and pitch angle of body are sequentially carried out asymmetric hybrid gait generation, speed consistency compensation, posture inverse kinematics compensation based on virtual plane and optimal distribution of anti-slip foot end force based on step-by-step quadratic programming;Step 3: verification and execution are carried out to step 2, realize the adaptive gait planning and posture control of hexapod robot under asymmetric narrow space constraint.The application is used to solve the problem of body tilting, foot slipping and internal force conflict faced by hexapod robot when walking in "rail-sleeper" asymmetric narrow space.

Description

Technical Field

[0001] This invention belongs to the field of motion control technology for multi-legged robots, specifically relating to an adaptive gait planning and posture control method for a six-legged robot based on asymmetric narrow space constraints. Background Technology

[0002] Hexapods are often used for complex terrain exploration due to their excellent stability. However, most existing hexapod gait planning methods (such as the standard triangular gait) assume that the terrain is homogeneous or continuous. In specific confined space scenarios such as railway inspection and tunnel surveying, robots face extreme asymmetric terrain constraints. The left constraint requires walking on smooth, hard rails with a certain height, where the coefficient of friction is low and slippage is easy. The right constraint requires walking on discrete, uneven sleepers and ballast, where footholds are limited and the height is lower than the rails.

[0003] Existing single-tripod gait or wave gait have significant drawbacks in this scenario: if a tripod gait is used, the right foot is prone to misstepping or the robot may tilt violently due to terrain elevation differences; if a pure wave gait is used, although stable, the movement speed is too slow to meet inspection efficiency requirements. In addition, the drastically different contact stiffness and friction characteristics of the left and right sides can easily cause internal force conflicts and rollover instability in the robot.

[0004] Therefore, there is an urgent need for an adaptive gait and control method that can adapt to both a continuous hard plane (railway) on one side and discrete unstructured terrain (sleepers) on the other side. Summary of the Invention

[0005] This invention provides an adaptive gait planning and attitude control method for a hexapod robot based on asymmetric narrow space constraints, in order to solve the problems of body tilting, foot slippage and internal force conflict faced by hexapod robots when walking in asymmetric narrow space such as "railway-sleeper".

[0006] This invention is achieved through the following technical solution: An adaptive gait planning and attitude control method for a hexapod robot based on asymmetric narrow spatial constraints, the method comprising the following steps: Step 1: Acquire environmental information and fuselage attitude, and perform asymmetric support surface modeling; Step 2: Based on the modeling in Step 1 and the roll and pitch angles of the robot body, the hexapod robot controller performs asymmetric hybrid gait generation, velocity consistency compensation, attitude inverse kinematics compensation based on the virtual plane, and optimal distribution of anti-slip foot end force based on stepwise quadratic programming. Step 3: Verify and execute Step 2 to achieve adaptive gait planning and attitude control of the hexapod robot under asymmetric narrow space constraints.

[0007] Furthermore, the asymmetric hybrid gait generation in step 2 specifically involves designing a split CPG network, where the leg on the left rail plane adopts a standard triangular gait, while the leg on the right sleeper discrete plane adopts a highly stable undulating gait, with the foot placement point forcibly locked in the central region of the sleeper, and a gait transformation factor is introduced. In the CPG model, the phase coupling relationship between the left and right oscillators is established using a gait transformation factor. Dynamically adjust the duty cycle on the left and right sides .

[0008] Furthermore, the speed consistency compensation in step 2 specifically involves all legs in a single gait cycle... Inside, the displacement of the support relative to the fuselage, i.e., the step length. They must be equal; set the target speed of the fuselage as... Then the step size on each side must satisfy: Desired stride length of the leg on the left rail plane:

[0009] Desired step size of the leg in the discrete plane of the right sleeper:

[0010] because The right leg needs to take a longer stride. ; For the leg on the discrete plane of the right sleeper, due to its longer support time, the swing time... The speed of the right swing leg is greatly compressed; the main control computing system uses fifth-order polynomial trajectory interpolation to automatically increase the instantaneous angular velocity of the right swing leg, ensuring that it can cross a larger stride in a shorter time and achieve vector alignment of the forward speed of the left and right legs.

[0011] Furthermore, step 2, the virtual plane-based attitude inverse kinematics compensation, specifically includes the following steps: Step B2.1: Establish a virtual horizontal reference plane with the higher left rail plane as the reference coordinate system. The sensing system measures in real time the height difference between the discrete plane of the right-side sleeper and the rail. ; Step B2.2: Attitude Error Mapping. The three-axis attitude angles are monitored in real time by the IMU built into the fuselage. When the fuselage tilts due to settlement or elevation difference on the right side of the terrain, the error is mapped. At that time, the controller performs speed compensation calculations and joint angle mapping calculations; Step B2.3: Based on the controller's calculations in step B2.2, perform real-time scaling compensation.

[0012] Furthermore, the velocity compensation calculation in step B2.2 specifically involves converting the attitude error into a corrected velocity for the fuselage center of gravity:

[0013] The desired linear velocity of the foot in the fuselage coordinate system :

[0014] in, For the first The position vector of the foot tip of a leg relative to the center of mass of the fuselage; The joint angle mapping calculation in step B2.2 is specifically as follows: Using the inverse kinematics formula based on the Jacobian matrix, the compensated velocities are mapped onto the 18 joints of the six legs: in For the Jacobian matrix of the robot, This refers to the angular velocity commands for each joint actuator.

[0015] Furthermore, step B2.3 specifically involves the right leg automatically landing at the target footing point. In addition to the above, add a length component. .

[0016] Furthermore, the optimal allocation of anti-slip foot force based on stepwise quadratic programming in step 2 specifically involves: Step U2.1: Establish an asymmetric friction cone constraint model. The contact force at the foot of the supporting leg is To prevent slippage, the force applied to the foot must be strictly within the friction cone of the local ground surface, i.e., satisfying the following:

[0017] in The coefficient of friction is the dynamic friction coefficient. The normal force must satisfy the maximum load-bearing constraint of a single leg: ; Step U2.2: First step in optimizing QP1: Optimize the foot force of all supporting legs. and the upper limit of the global traction coefficient ; The objective function is ; The resultant force and resultant torque generated by the robot's supporting legs must satisfy the desired six-degree-of-freedom motion of the robot body, expressed as:

[0018] in, For the overall quality of the robot; It is the vector of gravitational acceleration; Let be the desired linear acceleration vector of the robot's center of mass during the current control cycle; Let be the inertial tensor of the robot body in the center-of-mass coordinate system; This represents the current angular velocity vector of the robot body. Let be the desired angular acceleration vector of the robot body; Traction coefficient constraint:

[0019] The optimal set of decision variables is obtained through online optimization calculation using a quadratic programming solver. ; Step U2.3: Second step optimization QP2: The QP obtained in the first step Based on this, a second phase of planning will be conducted; the foot-end force of the supporting leg will be optimized. ; The objective function is , in This is the weight matrix; The result obtained in the first step Substitution makes It becomes a hard constraint; Solve for the optimal foot force Then, through the transpose of the Jacobian matrix... Mapped to the desired torque of each joint It is then sent down to the underlying motor driver for execution.

[0020] Furthermore, step 3 specifically involves using a three-dimensional quasi-static mechanical stability region as a safety criterion: In order to achieve optimal foot force Before the expected joint angle command is sent to the underlying motor, the main control computing system takes a step ahead and substitutes the current planning state into the three-dimensional static equilibrium equation. The main control computing system calculates whether the current center of gravity position is completely enveloped within the three-dimensional feasible polyhedron formed by the intersection of the friction cone boundaries of each supporting leg, the maximum torque boundary of the joint, and the normal support force boundary of the foot end in three-dimensional space. If the verification results show that the centroid is about to exceed the three-dimensional stability region, the controller immediately triggers the freeze mechanism: stopping the fall of the current oscillation phase and forcibly increasing the gait transformation factor. This extends the support phase time and re-triggers the secondary planning force allocation until the state returns to the stable domain.

[0021] Furthermore, step 3 specifically involves introducing position-based adaptive impedance control into the leg on the right side of the sleeper discrete plane; treating the leg on the right side of the sleeper discrete plane as an equivalent spring-damped system with mass, damping, and stiffness characteristics; and establishing a dynamic impedance relationship in the Z-axis direction at the foot end.

[0022] in These are the set target mass, target damping, and target stiffness parameters, respectively. The actual foot position, For reference location in planning; The vertical contact force is collected in real time by a one-dimensional force sensor at the foot. Based on Lyapunov's asymptotic stability theorem, an adaptive law is designed to estimate environmental stiffness online in real time. and environmental location :

[0023]

[0024] in, This is the predicted contact force value. Adaptive gain for positive constants; When the foot of the leg on the right side of the discrete plane of the sleeper steps on an unknown protruding sleeper or gravel, the impedance controller, combined with the online estimated environmental stiffness, triggers the circuit. Calculate a position correction amount .

[0025] An adaptive gait planning and attitude control method for hexapod robots based on asymmetric narrow space constraints, as described above, is applied to the adaptive gait generation and high-precision stability control of hexapod robots in asymmetric narrow structured environments such as railway inspection and tunnel surveying.

[0026] The beneficial effects of this invention are: The upper-level CPG gait planning and step-by-step QP solver of this invention incorporates body dynamics prediction, three-dimensional mechanical stability domain (3DQESR) and nonlinear friction cone constraints into the quadratic programming solution. This enables the robot to anticipate future states and optimize foot force distribution in advance, fundamentally eliminating the risk of rollover and slippage on slopes or abrupt terrain.

[0027] The robot of this invention can not only land accurately on asymmetrical and rough terrain, but also land lightly, which greatly reduces the damage of instantaneous impact to the motor reducer and body structure, extends the hardware life, and takes into account both global optimization and real-time response performance.

[0028] This invention can guarantee millisecond-level response to unknown disturbances in complex terrain, and also ensure the global optimality of the overall trajectory of the aircraft, demonstrating a strong adaptive adjustment capability in unknown environments.

[0029] This invention relies on foot force feedback and dynamic adjustment of impedance parameters, eliminating the need for centimeter-level precise modeling of terrain obstacles. Even in the event of visual perception errors, the underlying impedance logic can automatically compensate for height deviations through physical "compliance," significantly improving the robot's survivability in extreme unstructured environments such as ruins and rocks. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the structure of the present invention.

[0031] Figure 2 This is the robot gait phase diagram of the present invention.

[0032] Figure 3 This is a flowchart of the method of the present invention. Detailed Implementation

[0033] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of this application with unnecessary detail.

[0034] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0035] It should also be understood that the terminology used in this application specification is for the purpose of describing particular embodiments only and is not intended to limit the application. As used in this application specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0036] The following is in conjunction with the appendix to this application specification. Figure 1-3 The technical solutions in the embodiments of this application are clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0037] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the spirit of this application. Therefore, this application is not limited to the specific embodiments disclosed below.

[0038] Implementation Method 1 This embodiment provides an adaptive gait planning and attitude control method for a hexapod robot based on asymmetric narrow spatial constraints. The control system of the hexapod robot includes a first part, a sensing unit, consisting of a depth camera, a six-axis IMU, and foot force sensors; a second part, a main control computing unit, employing a high-performance embedded motherboard with a built-in quadratic programming (QP) solver and a CPG phase coupling algorithm module; and a third part, an execution unit, connected to each leg servo driver via a CAN bus or Ethernet, such as... Figure 1 As shown.

[0039] like Figure 3 As shown, the method includes the following steps: Step 1: Acquire environmental information and fuselage attitude, and perform asymmetric support surface modeling; Step 1 specifically involves obtaining the terrain point cloud ahead, fitting the railway track plane on the left and the sleeper discrete plane on the right using the least squares method, and calculating the height difference between the two planes. and the average spacing of the right-side sleepers .

[0040] Step 2: Based on the modeling in Step 1 and the roll and pitch angles of the robot body, the hexapod robot controller performs asymmetric hybrid gait generation, velocity consistency compensation, attitude inverse kinematics compensation based on the virtual plane, and optimal distribution of anti-slip foot end force based on stepwise quadratic programming. Step 3: Verify and execute Step 2 to achieve adaptive gait planning and attitude control of the hexapod robot under asymmetric narrow space constraints.

[0041] Step 3: In asymmetric, narrow spaces, due to the high uncertainty in the height and distribution of the right-side sleeper terrain, upper-level planning alone cannot completely prevent collisions or instability during execution. Therefore, this invention adds two safety mechanisms before the lower-level drive: "stability domain advance verification" and "contact moment compliant control".

[0042] Furthermore, the asymmetric hybrid gait generation in step 2 specifically involves designing a split CPG (Central Pattern Generator) network. Specifically, the leg on the left rail plane adopts a standard triangular gait, prioritizing speed and continuity, while the leg on the right sleeper discrete plane adopts a highly stable undulating gait, with the foot placement forcibly locked in the central region of the sleeper. A gait transformation factor is introduced. In the CPG model, the phase coupling relationship between the left and right oscillators is established to ensure that although the left and right legs have different gaits, their overall phase is coordinated and they do not "walk in the same direction." This is achieved through a gait transformation factor. Dynamically adjust the duty cycle on the left and right sides .

[0043] Left rail plane leg clearance ratio for: (Standard triangular gait, pursuing high-frequency inspection speed); Duty cycle of the leg of the right sleeper discrete plane for: (in This allows the leg on the right side of the sleeper discrete plane to enter a undulating gait, ensuring the static stability of the multi-leg support.

[0044] Furthermore, the speed consistency compensation in step 2 specifically involves all legs in a single gait cycle... Inside, the displacement of the support relative to the fuselage, i.e., the step length. They must be equal; set the target speed of the fuselage as... Then the step size on each side must satisfy: Desired stride length of the leg on the left rail plane:

[0045] Desired step size of the leg in the discrete plane of the right sleeper:

[0046] because The right leg needs to take a longer stride. ; For the leg on the discrete plane of the right sleeper, due to its longer support time, the swing time... The speed of the right swing leg is greatly compressed; the main control computing system uses fifth-order polynomial trajectory interpolation to automatically increase the instantaneous angular velocity of the right swing leg, ensuring that it can cross a larger stride in a shorter time and achieve vector alignment of the forward speed of the left and right legs.

[0047] Furthermore, step 2, the virtual plane-based attitude inverse kinematics compensation, specifically includes the following steps: To ensure that the fuselage remains "absolutely level" on uneven terrain, this invention employs closed-loop control based on IMU feedback and inverse velocity kinematics. Step B2.1: Establish a virtual horizontal reference plane with the higher left rail plane as the reference coordinate system. The sensing system measures in real time the height difference between the discrete plane of the right-side sleeper and the rail. ; Step B2.2: Attitude Error Mapping. The three-axis attitude angles are monitored in real time by the IMU built into the fuselage. When the fuselage tilts due to settlement or elevation difference on the right side of the terrain, the error is mapped. At that time, the controller performs speed compensation calculations and joint angle mapping calculations; Step B2.3: Based on the controller's calculations in step B2.2, perform real-time scaling compensation.

[0048] Furthermore, the velocity compensation calculation in step B2.2 specifically involves converting the attitude error into a corrected velocity for the fuselage center of gravity:

[0049] The desired linear velocity of the foot in the fuselage coordinate system :

[0050] in, For the first The position vector of the foot tip of a leg relative to the center of mass of the fuselage; The joint angle mapping calculation in step B2.2 is specifically as follows: Using the inverse kinematics formula based on the Jacobian matrix, the compensated velocities are mapped onto the 18 joints of the six legs: in For the Jacobian matrix of the robot, This refers to the angular velocity commands for each joint actuator.

[0051] Let the length of the robot's base segment be... The length of the segment is The length of the tibia is The three joint angles of a single leg are respectively , , To reduce the computing power overhead of the controller, this invention constructs an explicit resolution mechanism. Jacobian matrix ;matrix The elements are expanded as follows: The first row of elements (corresponding to the partial derivative of the X-axis velocity):

[0052]

[0053]

[0054] The second row of elements (corresponding to the partial derivative of the Y-axis velocity):

[0055]

[0056]

[0057] The third row of elements (corresponding to the partial derivative of the Z-axis velocity):

[0058]

[0059] .

[0060] Furthermore, step B2.3 specifically involves the right leg automatically landing at the target footing point. In addition to the above, add a length component. Effect: Even if the right sleeper is 20cm lower than the left rail, the right leg, through a "squat" or "extended" motion, compensates for the height difference, much like a car's hydraulic suspension, forcing the fuselage's center of gravity coordinates to be in the world coordinate system. The axis remains constant, achieving absolute horizontality.

[0061] Furthermore, the optimal distribution of anti-slip foot force based on stepwise quadratic programming in step 2 specifically involves, in an asymmetric narrow space, the left rail surface is smooth (friction coefficient...). (Extremely low), the surface of the right-side sleeper is rough (coefficient of friction) (Higher); if the conventional average force distribution algorithm is used, the left supporting leg is very likely to slip and spin due to the tangential force exceeding the friction limit; therefore, this invention proposes a stepwise quadratic programming foot end force distribution method that introduces asymmetric friction cone constraints; The nonlinear friction cone constraint is: .

[0062] in, The robot's first A supporting leg (e.g., the leg that touches the ground during a single gait cycle). Indicates the first The set of tangential components of the reaction force at the foot of each supporting leg in the contact plane represents the vector of horizontal frictional force at the foot. This represents the dynamic coefficient of friction between the foot and the local contact surface. In this asymmetric, confined space scenario, it is a variable that varies with the terrain (the smooth rail on the left has an extremely low coefficient). The rough sleeper on the right is relatively high. ); Indicates the first The component of the reaction force at the foot of the supporting leg in the normal direction (perpendicular to the ground) of the contact plane represents the vertical supporting force on the foot. Step U2.1: Establish an asymmetric friction cone constraint model. The contact force at the foot of the supporting leg is To prevent slippage, the force applied to the foot must be strictly within the friction cone of the local ground surface, i.e., satisfying the following:

[0063] in The coefficient of friction is the dynamic friction coefficient. The normal force must satisfy the maximum load-bearing constraint of a single leg: ; Step U2.2: Optimization of foot force distribution and upper limit of traction coefficient based on quadratic programming QP1. This step aims to obtain the most conservative global traction coefficient that satisfies dynamic equilibrium under asymmetric, confined space conditions through optimization calculations in the first step. The specific implementation process is as follows: 1) Construct the objective function: To minimize the upper limit of the global traction coefficient To optimize the objective, the objective function expression is as follows:

[0064] in, Let be the objective function (i.e., the optimization evaluation index) of the first quadratic programming problem described in this step. This represents the total number of supporting legs during the current gait cycle. It is the set of forces at the foot of all supporting legs; The global traction coefficient is to be optimized.

[0065] 2) Constructing constraints: (1) Dynamic equilibrium constraints: The resultant force and resultant torque generated by the robot's supporting legs must satisfy the desired six-degree-of-freedom motion of the robot body, expressed as:

[0066] in, For the overall quality of the robot; It is the vector of gravitational acceleration; Let be the desired linear acceleration vector of the robot's center of mass during the current control cycle; Let be the inertia tensor (inertia matrix) of the robot body in the center-of-mass coordinate system. This represents the current angular velocity vector of the robot body. Let be the desired angular acceleration vector of the robot body.

[0067] (2) Traction coefficient (friction cone) constraint: To prevent the feet from slipping on the rails or sleepers, the force applied to the feet of each supporting leg must fall within the friction cone area:

[0068] in This is the collection of the tangential components of the foot reaction force within the contact plane, representing the horizontal frictional force. This represents the component of the foot reaction force in the normal direction of the contact plane, and signifies the vertical support force. Solution results: The optimal set of decision variables was obtained through online optimization calculation using a quadratic programming solver. ;in This is the upper limit of the most conservative traction coefficient to ensure that slippage does not occur under the current terrain.

[0069] Step U2.3: Second step optimization QP2: Minimize joint energy consumption and internal force resistance. Although the first step optimization solved the slippage problem, it may lead to an infinite number of solutions or cause excessive joint output torque; therefore, the optimization obtained in the first step... Based on this, a second phase of planning will be conducted; the foot-end force of the supporting leg will be optimized. ; The objective function is , in The weight matrix is ​​designed to minimize the magnitude of the resultant force at the foot end, thereby reducing the energy consumption of the joint motor. Maintain the dynamic equilibrium constraints from the first step; apply a strictly tightened friction cone constraint: [This is the result obtained in the first step]. Substitution makes It becomes a hard constraint; Solve for the optimal foot force Then, through the transpose of the Jacobian matrix... Mapped to the desired torque of each joint It is then sent down to the underlying motor driver for execution.

[0070] Furthermore, step 3 specifically involves verifying the three-dimensional mechanical stability region (3DFSR) before the action is executed. Traditional two-dimensional centroid projection polygons (such as ZMP) cannot reflect the true stability under asymmetric elevation differences; this invention uses a three-dimensional quasi-static mechanical stability region (3DQESR) as the safety criterion. In order to achieve optimal foot force Before the expected joint angle command is issued to the underlying motor of the execution and drive unit, the main control computing system takes a step ahead and substitutes the current planning state into the three-dimensional static equilibrium equation. The main control computing system calculates whether the current center of gravity position is completely enveloped within the three-dimensional feasible polyhedron (3DFSR) formed by the intersection of the friction cone boundaries of each supporting leg, the maximum moment boundary of the joint, and the normal support force boundary of the foot end in three-dimensional space; If the verification results show that the centroid is about to exceed the three-dimensional stability region (indicating a risk of rollover or slippage), the controller immediately triggers the freeze mechanism: stopping the descent of the current oscillation phase and forcibly increasing the gait transformation factor. This extends the support phase time and re-triggers the secondary planning force allocation until the state returns to the stable domain.

[0071] Furthermore, step 3 specifically involves the following: based on foot force feedback and Lyapunov adaptive compliant landing control, at the moment the right foot (sleeper side) transitions from the swing phase to the support phase and contacts the ground, if premature ground contact occurs due to terrain protrusion, the rigid position control will generate a huge contact impact force, causing fuselage vibration. This invention introduces position-based adaptive impedance control (Impedance Control) on the leg in the discrete plane of the right sleeper. Impedance model establishment: The leg in the discrete plane of the right sleeper is equivalent to a spring-damped system with mass, damping, and stiffness characteristics. An impedance dynamic relationship is established in the Z-axis direction of the foot.

[0072] in These are the set target mass, target damping, and target stiffness parameters, respectively. The actual foot position, For reference location in planning; The vertical contact force is collected in real time by a one-dimensional force sensor at the foot. Based on Lyapunov's indirect adaptive law: taking into account the stiffness of the sleepers and gravel (environmental stiffness) The environmental stiffness is unknown and constantly changing; therefore, this system employs an indirect adaptive control algorithm. Based on Lyapunov's asymptotic stability theorem, an adaptive law is designed to estimate the environmental stiffness online in real time. and environmental location :

[0073]

[0074] in, This is the predicted contact force value. Adaptive gain for positive constants; When the foot of the leg on the right side of the discrete plane of the sleeper steps on an unknown protruding sleeper or gravel, the impedance controller, combined with the online estimated environmental stiffness, triggers the circuit. It will quickly calculate a position correction amount. ; When the swing leg contacts an unmodeled protruding obstacle, the leg on the right sleeper discrete plane will actively and compliantly retract like an advanced independent suspension spring, absorbing the impact energy and dissipating excess impact force. Converted to foot position correction amount This allows the robot to not only land accurately but also land lightly on asymmetrical, rough terrain. It significantly reduces the damage to the motor reducer and chassis structure caused by instantaneous impacts, extending hardware lifespan and balancing global optimization with real-time response performance. By decoupling the control frequency: the upper-level CPG gait planning and step-by-step QP solver of this invention perform global optimal pose prediction at low frequencies (e.g., 50Hz), while the lower-level PID performs motor smooth following at high frequencies (e.g., 1000Hz). This "separation of light and heavy" calculation strategy ensures both millisecond-level response to unknown disturbances in complex terrain and global optimality of the overall aircraft trajectory, resulting in extremely strong adaptive adjustment capabilities in unknown environments.

[0075] The structure of the hexapod robot in this embodiment is as follows: Figure 1 As shown, it includes 6 legs, each of which includes a knee joint, a hip joint, and a root joint from bottom to top. The root joint is connected to the robot body, and a vision sensor (depth camera) is set at the front of the robot body.

[0076] Figure 2 for Figure 1 Gait phase diagram of a hexapod robot.

[0077] like Figure 1 and Figure 2 As shown in the figure, this embodiment simulates a robot inspecting a conveyor belt in the main transport roadway of a coal mine in a damp coal mine. Specifically, the robot performs the following inspection: 1. Environmental parameters and robot specifications: Left-side terrain (railway): P60 heavy steel rails, height .

[0078] Topography on the right (edge ​​of the drainage ditch): discrete concrete slabs and gravel pits. Perceptual modeling shows the right-side plane is lower than the left-side railroad track. (Exceeding the height of the robot's base segment). Robot status: The IMU detected the robot body tilting to the right at an angle. .

[0079] 2. Kinematic self-verification Inverse kinematic posture compensation: The controller initiates a "squat" mechanism. The left rail is used as the reference coordinate system. Calculate the extra extension of the right leg. By analyzing the Jacobian matrix Velocity inverse mapping The compensation amount is precisely distributed to the 18 joint angles of the six legs. This allows the fuselage to return to a horizontal position within 1.5 seconds, reducing the center of gravity height deviation to [missing value]. Within.

[0080] 3.3DFSR verification and compliant landing: Safety Verification: Before stepping onto the next cover plate, the system calculates the 3DQESR stability region. If the landing point is found to be in a loose gravel area, the stability region boundary will shrink sharply, and the system will immediately trigger a "circuit breaker" to adjust the gait transformation factor. To increase the support coverage area.

[0081] Compliant landing: At the moment of contact between the right forefoot and the concrete cover, the environmental stiffness is estimated in real time using Lyapunov's adaptive law. When the sensor detects When subjected to instantaneous impact force, the impedance controller actively generates The retraction displacement absorbed The collision energy prevents the fuselage from shaking when there are drastic changes in altitude.

[0082] Implementation Method 2 This embodiment describes an adaptive gait planning and attitude control method for a six-legged robot based on asymmetric narrow space constraints, as described in Embodiment 1. It is applied to the adaptive gait generation and high-precision stability control of a six-legged robot in asymmetric narrow structured environments such as railway inspection and tunnel surveying.

Claims

1. A method for adaptive gait planning and attitude control of a hexapod robot based on asymmetric narrow spatial constraints, characterized in that, The method includes the following steps: Step 1: Acquire environmental information and fuselage attitude, and perform asymmetric support surface modeling; Step 2: Based on the modeling in Step 1 and the roll and pitch angles of the robot body, the hexapod robot controller performs asymmetric hybrid gait generation, velocity consistency compensation, attitude inverse kinematics compensation based on the virtual plane, and optimal distribution of anti-slip foot end force based on stepwise quadratic programming. Step 3: Verify and execute Step 2 to achieve adaptive gait planning and posture control of the hexapod robot under asymmetric narrow space constraints.

2. The method according to claim 1, characterized in that, The asymmetric hybrid gait generation in step 2 specifically involves designing a split CPG network, where the legs on the left rail plane adopt a standard triangular gait, while the legs on the right sleeper discrete plane adopt a highly stable undulating gait, with the foot placement point forcibly locked in the central region of the sleeper, and a gait transformation factor is introduced. In the CPG model, the phase coupling relationship between the left and right oscillators is established using a gait transformation factor. Dynamically adjust the duty cycle on the left and right sides .

3. The method according to claim 1, characterized in that, The speed consistency compensation in step 2 specifically involves all legs maintaining speed consistency during a single gait cycle. Inside, the displacement of the support relative to the fuselage, i.e., the step length. They must be equal; set the target speed of the fuselage as... Then the step size on each side must satisfy: Desired stride length of the leg on the left rail plane: Desired step size of the leg in the discrete plane of the right sleeper: because The right leg needs to take a longer stride. ; For the leg on the discrete plane of the right sleeper, due to its longer support time, the swing time... The speed of the right swing leg is greatly compressed; the main control computing system uses fifth-order polynomial trajectory interpolation to automatically increase the instantaneous angular velocity of the right swing leg, ensuring that it can cross a larger stride in a shorter time and achieve vector alignment of the forward speed of the left and right legs.

4. The method according to claim 1, characterized in that, Step 2, the virtual plane-based attitude inverse kinematics compensation, specifically includes the following steps: Step B2.1: Establish a virtual horizontal reference plane with the higher left rail plane as the reference coordinate system. The sensing system measures in real time the height difference between the discrete plane of the right-side sleeper and the rail. ; Step B2.2: Attitude Error Mapping. The three-axis attitude angles are monitored in real time by the IMU built into the fuselage. When the fuselage tilts due to settlement or elevation difference on the right side of the terrain, the error is mapped. At that time, the controller performs speed compensation calculations and joint angle mapping calculations; Step B2.3: Based on the controller's calculations in step B2.2, perform real-time scaling compensation.

5. The method according to claim 4, characterized in that, The speed compensation calculation in step B2.2 specifically involves converting the attitude error into a corrected speed for the fuselage center of gravity: The desired linear velocity of the foot in the fuselage coordinate system : in, For the first The position vector of the foot tip of a leg relative to the center of mass of the fuselage; The joint angle mapping calculation in step B2.2 is specifically as follows: Using the inverse kinematics formula based on the Jacobian matrix, the compensated velocities are mapped onto the 18 joints of the six legs: in For the Jacobian matrix of the robot, This refers to the angular velocity commands for each joint actuator.

6. The method according to claim 4, characterized in that, Specifically, step B2.3 involves the right leg automatically landing at the target footing point. In addition to the above, add a length component. .

7. The method according to claim 1, characterized in that, The optimal allocation of anti-slip foot force based on stepwise quadratic programming in step 2 is specifically as follows: Step U2.1: Establish an asymmetric friction cone constraint model. The contact force at the foot of the supporting leg is To prevent slippage, the force applied to the foot must be strictly within the friction cone of the local ground surface, i.e., satisfying the following: in The coefficient of friction is the dynamic friction coefficient. The normal force must satisfy the maximum load-bearing constraint of a single leg: ； Step U2.2: First step in optimizing QP1: Optimize the foot force of all supporting legs. and the upper limit of the global traction coefficient ; The objective function is ; The resultant force and resultant torque generated by the robot's supporting legs must satisfy the desired six-degree-of-freedom motion of the robot body, expressed as: in, For the overall quality of the robot; It is the vector of gravitational acceleration; Let be the desired linear acceleration vector of the robot's center of mass during the current control cycle; Let be the inertial tensor of the robot body in the center-of-mass coordinate system; This represents the current angular velocity vector of the robot body. Let be the desired angular acceleration vector of the robot body; Traction coefficient constraint: The optimal set of decision variables is obtained through online optimization calculation using a quadratic programming solver. ; Step U2.3: Second step optimization QP2: The QP obtained in the first step Based on this, a second phase of planning will be conducted; the foot-end force of the supporting leg will be optimized. ; The objective function is , in This is the weight matrix; The result obtained in the first step Substitution makes It becomes a hard constraint; Solve for the optimal foot force Then, through the transpose of the Jacobian matrix... Mapped to the desired torque of each joint It is then sent down to the underlying motor driver for execution.

8. The method according to claim 1, characterized in that, Specifically, step 3 involves using a three-dimensional quasi-static mechanical stability region as a safety criterion. In order to achieve optimal foot force Before the expected joint angle command is sent to the underlying motor, the main control computing system takes a step ahead and substitutes the current planning state into the three-dimensional static equilibrium equation. The main control computing system calculates whether the current center of gravity position is completely enveloped within the three-dimensional feasible polyhedron formed by the intersection of the friction cone boundaries of each supporting leg, the maximum torque boundary of the joint, and the normal support force boundary of the foot end in three-dimensional space. If the verification results show that the centroid is about to exceed the three-dimensional stability region, the controller immediately triggers the freeze mechanism: stopping the fall of the current oscillation phase and forcibly increasing the gait transformation factor. This extends the support phase time and re-triggers the secondary planning force allocation until the state returns to the stable domain.

9. The method according to claim 1, characterized in that, Specifically, step 3 involves introducing position-based adaptive impedance control into the leg on the right side of the sleeper discrete plane; treating the leg on the right side of the sleeper discrete plane as an equivalent spring-damped system with mass, damping, and stiffness characteristics; and establishing a dynamic impedance relationship in the Z-axis direction at the foot end. in These are the set target mass, target damping, and target stiffness parameters, respectively. The actual foot position, For reference location in planning; The vertical contact force is collected in real time by a one-dimensional force sensor at the foot. Based on Lyapunov's asymptotic stability theorem, an adaptive law is designed to estimate environmental stiffness online in real time. and environmental location : in, This is the predicted contact force value. Adaptive gain for positive constants; When the foot of the leg on the right side of the discrete plane of the sleeper steps on an unknown protruding sleeper or gravel, the impedance controller, combined with the online estimated environmental stiffness, triggers the circuit. Calculate a position correction amount .

10. A method for adaptive gait planning and attitude control of a hexapod robot based on asymmetric narrow space constraints as described in any one of claims 1-9, which is applied to adaptive gait generation and high-precision stability control of hexapod robots in asymmetric narrow structured environments such as railway inspection and tunnel surveying.

Images