Task-adaptive whole-body variable impedance control method for wheeled humanoid robot

By fusing multi-source heterogeneous signals and task-dependent impedance scheduling, the system achieves full-body variable impedance control for wheeled humanoid robots, resolving the contradiction between assembly compliance and handling stability, improving the system's task adaptability and safety, and making it suitable for complex and ever-changing task scenarios.

CN121608155APending Publication Date: 2026-03-06WANJING QIANXUN (BEIJING) TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202512044646.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-12-31
Publication Date
2026-03-06

AI Technical Summary

Technical Problem

Existing control methods for wheeled humanoid robots struggle to balance compliance during assembly and stability during transport, and lack the ability to optimize whole-body coordination. They are unable to effectively address complex and ever-changing task requirements, especially in hybrid tasks involving movement and operation, where problems such as rigid collisions, posture instability, and wheel-to-ground slippage exist.

Method used

The system employs a method of real-time fusion estimation of multi-source heterogeneous signals, adaptive scheduling of task-dependent impedance parameters, upper body impedance control and chassis virtual force admittance mapping. By adjusting the upper body stiffness and chassis damping in real time through slip ratio, contact stiffness and attitude disturbance, the system achieves full-body variable impedance control, ensuring system stability. The control logic is executed at a frequency of 1kHz on the x86 real-time computing platform.

Benefits of technology

It achieves a balance between assembly compliance and handling stability, supports hybrid tasks that involve moving and operating simultaneously, improves the system's task generalization and operational safety, and is suitable for application scenarios such as 3C assembly, logistics handling, and service interaction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121608155A_ABST
    Figure CN121608155A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of wheeled humanoid robot control, in particular to a task adaptive whole-body variable impedance control method for a wheeled humanoid robot, which comprises the steps of multi-source signal fusion estimation, task dependent impedance scheduling, upper body impedance control, chassis virtual force admittance mapping and 1kHz real-time closed-loop execution. The upper body rigidity and the chassis damping can be adjusted in real time based on the slip rate, the contact rigidity and the attitude disturbance, on the premise that the passivity and the stability of the system are guaranteed, unification of the assembling flexibility and the carrying stability is achieved, and the mixed task of moving and operating at the same time is supported.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of robot control technology, and in particular to a task-adaptive whole-body variable impedance control method for a wheeled humanoid robot. Background Technology

[0002] Wheeled humanoid robots, as highly flexible intelligent systems integrating a mobile chassis and a humanoid upper body structure, are increasingly being applied to complex scenarios such as industrial assembly, logistics handling, service interaction, and special operations. In these applications, robots frequently need to switch between task types. For example, in precision assembly tasks, the arm must have high compliance to absorb positional errors and avoid excessive contact force that could damage the workpiece, while the chassis must remain stable. In rapid handling tasks, the arm must have high rigidity to hold the object firmly, and the chassis must have strong anti-disturbance capabilities to cope with uneven ground or inertial impacts. However, existing control methods generally use fixed impedance parameters (such as constant stiffness K and damping D), making it difficult to balance these conflicting mechanical interaction requirements. This leads to rigid collisions during assembly and instability or even wheel slippage during handling.

[0003] More critically, current control architectures for wheeled humanoid robots typically treat the chassis and upper body as separate entities: the chassis often uses a speed control interface, while the arms rely on torque control to achieve impedance behavior. The two lack a unified mechanical interaction framework, hindering full-body coordinated optimization. This control heterogeneity severely limits the system's adaptability to dynamic environments, especially in hybrid tasks involving movement and manipulation. Furthermore, existing solutions generally lack deep perception and feedback mechanisms for environmental states, failing to effectively utilize multi-source heterogeneous information such as wheel-to-ground slip ratio, end-effector contact stiffness, and body posture disturbances for real-time adaptive adjustment. This results in rigid control strategies that struggle to meet complex and ever-changing task requirements.

[0004] Meanwhile, even when some studies attempt to introduce variable impedance control, they often neglect system stability assurance during online parameter adjustment, posing a risk of oscillations or even divergence due to sudden parameter changes. Furthermore, wheeled humanoid robots inherently possess high-dimensional nonlinear dynamics; achieving multi-sensor fusion, parameter scheduling, and command generation within high-frequency control cycles above 1 kHz requires resolving the trade-off between computational efficiency and real-time performance. Therefore, there is an urgent need for a whole-body variable impedance control method that can incorporate a speed-controlled chassis into a unified impedance control framework, fuse multimodal environmental perception signals, achieve task adaptation, and possess strict stability guarantees. This method would overcome the bottlenecks of existing technologies in terms of task generalization, collaborative consistency, and operational safety. Summary of the Invention

[0005] The problem solved by this invention is to provide a task-adaptive whole-body variable impedance control method for a wheeled humanoid robot, which can adjust the upper body stiffness and chassis damping in real time based on slip ratio, contact stiffness and posture disturbance, so as to achieve the unity of assembly compliance and handling stability while ensuring the passivity and stability of the system, and support hybrid tasks of moving and operating at the same time.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a task-adaptive whole-body variable impedance control method for a wheeled humanoid robot, comprising the following steps:

[0007] S1. Real-time fusion estimation of multi-source heterogeneous signals;

[0008] S2, Adaptive scheduling based on task-dependent impedance parameters;

[0009] S3, Upper body impedance control is achieved;

[0010] S4, Chassis Virtual Force Admittance Mapping;

[0011] S5, Real-time closed-loop execution;

[0012] In step S1, a three-signal fusion estimator synchronously processes raw data from the wheel speed encoder, joint torque sensor, and six-axis IMU to generate three types of state signals: slip ratio, contact stiffness, and attitude disturbance. Step S2, based on these three state signals, generates time-varying upper body stiffness, chassis damping, and chassis stiffness using a task-dependent impedance scheduler, and applies triple stability guarantees: parameter change rate limits, critical damping constraints, and virtual damping injection. Step S3 implements impedance control in the upper body operating space, mapping the desired interaction force into joint torque commands through a Jacobi transpose mapping module. Step S4 constructs virtual forces in the chassis motion space and maps them into speed commands through an admittance model module, then generates wheel speed commands through an omnidirectional wheel inverse kinematics solution module. Step S5 executes all control logic cyclically at a 1kHz control cycle on an x86 real-time computing platform.

[0013] Preferably, step S1 includes the following three parallel sub-processes:

[0014] (1) The slip ratio estimation unit is based on the wheel speed v output by the wheel speed encoder. wheel The linear velocity v of the body measured by the IMU body Calculate the slip ratio Where ∈=10 -3 m / s, and the result is low-pass filtered with a cutoff frequency of 10Hz;

[0015] (2) The contact stiffness identification unit updates the equivalent contact stiffness k online based on the end force FF measured by the joint torque sensor and the end pose increment Δx obtained by kinematic calculation. c Updates are activated only when the end force amplitude is greater than 5N and the displacement increment is greater than 0.1mm.

[0016] (3) The attitude disturbance detection unit calculates the disturbance measure d = ||ω||² + ||ag||² based on the angular velocity vector ω and linear acceleration vector a output by the IMU, where the gravity vector is...

[0017] Preferably, the covariance matrix update formula of the recursive least squares algorithm is as follows:

[0018] The regression vector φ(k) = Δx(k) has a forgetting factor λ = 0.98, and the sliding window length is fixed at 100 sampling points.

[0019] Preferably, in step S2, the upper body stiffness K u (t) is determined by the assembly compliance function module based on the contact stiffness k. c Confirmed, the expression is:

[0020]

[0021] Among them, the high stiffness threshold k high =5000N / m.

[0022] Preferably, in step S2, the chassis damping D b (t) is determined by the transport steady-state function module based on the slip ratio ss and the disturbance measure dd, and the expression is: D b (t)=50+250·min(1,β1s+β2d),

[0023] The weighting coefficients β1 = 4.0 s and β2 = 0.2 s / rad, and D b (t)∈[50,300]Ns / m.

[0024] Preferably, in step S2, the chassis stiffness K b (t) is determined by the anti-slip enhancement function module based on the slip ratio s, and the expression is: K b (t)=500+γ·s

[0025] , where the gain coefficient γ = 2000N / (m\cdotps).

[0026] Preferably, the triple stability guarantee applied in step S2 includes:

[0027] The parameter change rate limiter applies a first-order differential constraint to the upper body stiffness and chassis damping, i.e.

[0028] The critical damping constraint check module requires that the actual damping meets the following requirements. Where the damping ratio ζ = 0.7, and the upper body equivalent mass M u =10kg, chassis equivalent mass M b =80kg;

[0029] The virtual damping injection module introduces a fixed D in the admittance model. adm Virtual damping term = 20 Ns / m.

[0030] Preferably, in step S3, the upper body impedance control branch calculates the desired interaction force: Among them, upper body damping And generate the joint torque command τ through the Jacobi transpose mapping module. arm =J T (q)F des The torque is then transmitted to the torque controllers of the arms and waist / leg joints.

[0031] Preferably, in step S4, the chassis virtual force admittance mapping branch constructs a virtual force: F virtual =K b (t)e p +D b (t)e v The position error e p =p d -p, speed error The virtual force is then input into the admittance model module to generate a velocity command. The compensation speed output by the superimposed slip compensation module Obtain the final speed command The omnidirectional wheel inverse kinematics calculation module then calculates the angular velocity commands for the four wheels.

[0032] Preferably, step S5 is deployed on an x86 industrial computer platform, with the operating system using the Ubuntu 20.04LTS kernel and Xenomai real-time extension. The main control thread is bound to a dedicated CPU core and configured with the SCHED_FIFO scheduling strategy. The control flow includes five functional modules: sensor reading, state estimation, parameter scheduling, dynamic mapping, and instruction issuance. It employs batch reading of sensor data, cached reuse of the Jacobian matrix, Cholesky pre-decomposition to accelerate the RLS algorithm, lookup table method combined with linear interpolation to implement the assembly compliance function, analytical expression to calculate the Jacobian matrix, pre-stored omnidirectional wheel configuration matrix, single-precision floating-point arithmetic, and EtherCAT distributed clock synchronization technology to ensure stable execution of the 1ms control cycle.

[0033] The beneficial effects of this invention are as follows: Through the three-signal fusion estimator in step S1, the system can perceive three key environmental states in real time: wheel-ground slippage, hand-end contact stiffness, and body posture disturbance, providing reliable input for parameter adaptation; through the task-dependent scheduling mechanism in step S2, the upper body stiffness is continuously adjusted within the range of 150–2000 N / m, and the chassis damping is dynamically enhanced within the range of 50–300 Ns / m, achieving a balance between assembly compliance and handling stability; through the chassis virtual force admittance mapping in step S4, the originally isolated speed control chassis is incorporated into the whole-body impedance framework, making the chassis exhibit equivalent impedance port characteristics externally, forming a consistent mechanical interaction behavior with the upper body; through the triple stability constraints in S2, the system always satisfies the critical damping condition and passivity during online parameter changes, avoiding oscillation or divergence; through the real-time optimization technology in S5, the entire method runs stably at a frequency of 1 kHz on a standard industrial PC, supporting mixed task scenarios of simultaneous movement and precision operation, and is suitable for various applications such as 3C assembly, logistics handling, and service interaction. Attached Figure Description

[0034] Figure 1 This is a diagram illustrating the overall system architecture of the task-adaptive whole-body variable impedance control method for the wheeled humanoid robot of the present invention.

[0035] Figure 2 This is a schematic diagram of the structure and signal flow of the impedance control branch of the present invention;

[0036] Figure 3 This is a schematic diagram of the structure and signal flow of the chassis virtual force admittance mapping branch of the present invention. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] Specific implementation examples are given below.

[0039] See Figures 1-3 A task-adaptive whole-body variable impedance control method for a wheeled humanoid robot includes the following steps:

[0040] S1. Real-time fusion estimation of multi-source heterogeneous signals;

[0041] S2, Adaptive scheduling based on task-dependent impedance parameters;

[0042] S3, Upper body impedance control is achieved;

[0043] S4, Chassis Virtual Force Admittance Mapping;

[0044] S5, Real-time closed-loop execution;

[0045] In step S1, a three-signal fusion estimator synchronously processes raw data from the wheel speed encoder, joint torque sensor, and six-axis IMU to generate three types of state signals: slip ratio, contact stiffness, and attitude disturbance. Step S2, based on these three state signals, generates time-varying upper body stiffness, chassis damping, and chassis stiffness using a task-dependent impedance scheduler, and applies triple stability guarantees: parameter change rate limits, critical damping constraints, and virtual damping injection. Step S3 implements impedance control in the upper body operating space, mapping the desired interaction force into joint torque commands through a Jacobi transpose mapping module. Step S4 constructs virtual forces in the chassis motion space and maps them into speed commands through an admittance model module, then generates wheel speed commands through an omnidirectional wheel inverse kinematics solution module. Step S5 executes all control logic cyclically at a 1kHz control cycle on an x86 real-time computing platform.

[0046] The S1 step includes the following three parallel sub-processes:

[0047] (1) The slip ratio estimation unit is based on the wheel speed v output by the wheel speed encoder. wheel The linear velocity v of the body measured by the IMU body Calculate the slip ratio Where ∈=10 -3 m / s, and the result is low-pass filtered with a cutoff frequency of 10Hz;

[0048] (2) The contact stiffness identification unit updates the equivalent contact stiffness k online based on the end force FF measured by the joint torque sensor and the end pose increment Δx obtained by kinematic calculation. c Updates are activated only when the end force amplitude is greater than 5N and the displacement increment is greater than 0.1mm.

[0049] (3) The attitude disturbance detection unit calculates the disturbance measure d = ||ω||² + ||ag||² based on the angular velocity vector ω and linear acceleration vector a output by the IMU, where the gravity vector is...

[0050] The covariance matrix update formula of the recursive least squares algorithm is as follows:

[0051] Where the regression vector φ(k) = Δx(k), the forgetting factor λ = 0.98, and the sliding window length is fixed at 100 sampling points; in step S2, the upper body stiffness K u (t) is determined by the assembly compliance function module based on the contact stiffness k. c Confirmed, the expression is:

[0052]

[0053] Among them, the high stiffness threshold k high =5000N / m; In step S2, the chassis damping D b (t) is determined by the transport steady-state function module based on the slip ratio ss and the disturbance measure dd, and the expression is: D b (t)=50+250·min(1,β1s+β2d),

[0054] The weighting coefficients β1 = 4.0 s and β2 = 0.2 s / rad, and D b (t)∈[50,300]Ns / m;

[0055] In step S2, the chassis stiffness K b (t) is determined by the anti-slip enhancement function module based on the slip ratio s, and the expression is: K b (t)=500+γ·s

[0056] Where the gain coefficient γ = 2000 N / (m\cdotps);

[0057] The triple stability guarantee applied in step S2 includes:

[0058] The parameter change rate limiter applies a first-order differential constraint to the upper body stiffness and chassis damping, i.e.

[0059] The critical damping constraint check module requires that the actual damping meets the following requirements. Where the damping ratio ζ = 0.7, and the upper body equivalent mass M u =10kg, chassis equivalent mass M b =80kg;

[0060] The virtual damping injection module introduces a fixed D in the admittance model. adm =20 Ns / m virtual damping term; in step S3, the upper body impedance control branch calculates the expected interaction force: Among them, upper body damping And generate the joint torque command τ through the Jacobi transpose mapping module. arm =J T (q)F des The torque is then transmitted to the torque controllers of the arms and the waist and leg joints;

[0061] In step S4, the chassis virtual force admittance mapping branch constructs the virtual force: F virtual =K b (t)e p +D b (t) xv The position error e p =p d -p, speed error The virtual force is then input into the admittance model module to generate a velocity command. The compensation speed output by the superimposed slip compensation module Obtain the final speed command The omnidirectional wheel inverse kinematics calculation module then calculates the angular velocity commands for the four wheels.

[0062] The S5 step is deployed on an x86 industrial computer platform, using the Ubuntu 20.04 LTS kernel with Xenomai real-time extensions. The main control thread is bound to a dedicated CPU core and configured with the SCHED_FIFO scheduling strategy. The control flow includes five functional modules: sensor reading, state estimation, parameter scheduling, dynamic mapping, and command issuance. It employs batch reading of sensor data, buffering and reusing the Jacobian matrix, Cholesky pre-decomposition to accelerate the RLS algorithm, using lookup table method combined with linear interpolation to implement the assembly compliance function, calculating the Jacobian matrix using analytical expressions, pre-storing the omnidirectional wheel configuration matrix, single-precision floating-point arithmetic, and EtherCAT distributed clock synchronization technology to ensure stable execution of the 1ms control cycle.

[0063] The modular architecture fully realizes closed-loop control from environmental perception, parameter scheduling, mechanical mapping to real-time execution at a control frequency of 1kHz, solving the core problems faced by wheeled humanoid robots in multi-task scenarios, such as the contradiction between compliance and stability, control heterogeneity and stability risks.

[0064] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for task-adaptive whole-body variable impedance control of a wheeled humanoid robot, characterized by, The method comprises the following steps: S1, real-time fusion estimation of multi-source heterogeneous signals; S2, task-dependent impedance parameter adaptive scheduling; S3, upper body impedance control implementation; S4, chassis virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual force and virtual ​ ​ ​ 2. The task-adaptive whole-body variable impedance control method of the wheeled humanoid robot according to claim 1, wherein, ​ (1) The slip ratio estimation unit calculates the slip ratio wheel from the wheel speed v body measured by the IMU where ∈ = 10 -3 m / s and low-pass filters the result with a cutoff frequency of 10 Hz; (2) The contact stiffness identification unit uses the end force FF measured by the joint torque sensor and the end pose increment Δx obtained by kinematics calculation to update the equivalent contact stiffness k online using the recursive least squares algorithm c The update is activated only when the end force amplitude is greater than 5 N and the displacement increment is greater than 0.1 mm. (3) The attitude disturbance detection unit computes a disturbance measure d = ||ω||2+ ||a - g||2 based on the angular velocity vector ω and the linear acceleration vector a output by the IMU, where the gravity vector g = (0, 0, g) is known.

3. The task-adaptive whole-body variable impedance control method of the wheeled humanoid robot according to claim 2, wherein, The covariance matrix update formula of the recursive least squares algorithm is ​ 4. The task-adaptive whole-body variable impedance control method of the wheeled humanoid robot according to claim 1, wherein, The upper body rigidity K u (t) is determined by the assembly compliance function module according to the contact rigidity k c The expression is: where the high stiffness threshold k high = 5000 N / m. 5.The task-adaptive whole-body variable impedance control method of the wheeled humanoid robot according to claim 1, wherein, In said S2 step, the chassis damping D b (t) is determined by the transport steady function module as a function of the slip rate ss and the disturbance measure dd, expressed as: D b (t) = 50 + 250 · min(1, β1s + β2d), where the weighting coefficients are β1= 4.0 s, β2= 0.2 s / rad, and D b (t) ∈ [50, 300] Ns / m. 6.The task-adaptive whole-body variable impedance control method of the wheeled humanoid robot according to claim 1, wherein, In the S2 step, the chassis stiffness K b (t) is determined by the anti-slip enhancement function module according to the slip rate s, and the expression is: K b (t) = 500 + γ · s ​ 7.The task-adaptive whole-body variable impedance control method of the wheeled humanoid robot according to claim 1, wherein, ​ The parameter rate limiter imposes a first-order differential constraint on the upper body stiffness and chassis damping, i.e. The critical damping constraint check module requires that the actual damping satisfies where the damping ratio ζ = 0.7, the upper body equivalent mass M u = 10 kg, and the chassis equivalent mass M b = 80 kg. The virtual damping injection module fixes the introduction of D in the admittance model adm = 20 Ns / m virtual damping term. 8.The task-adaptive whole-body variable impedance control method of the wheeled humanoid robot according to claim 1, wherein, In the S3 step, the upper body impedance control branch calculates the expected interaction force: where the upper body damping and generates joint torque commands τ arm = J T (q) F des to the dual-arm and waist / leg joint torque controllers. 9.The task-adaptive whole-body variable impedance control method of the wheeled humanoid robot according to claim 1, wherein, The S4 step, chassis virtual force and virtual force and virtual force virtual = K b (t) e p + D b (t) e v , where the position error e p = p d -p, the velocity error and input the virtual force into the virtual force and virtual force The compensation velocity output by the slip compensation module is superimposed The final velocity command is obtained And the omni-directional wheel inverse kinematics solution module is solved into four-wheel angular velocity command. 10.The task-adaptive whole-body variable impedance control method of the wheeled humanoid robot according to claim 1, wherein, ​