Abstract:Gait control of legged robotic walkers on dynamically moving surfaces (e.g., ships and vehicles) is challenging due to the limited balance control actuation and unknown surface motion. We present a contingent model predictive control (CMPC) for bipedal walker locomotion on moving surfaces with a linear inverted pendulum (LIP) model. The CMPC is a robust design that is built on regular model predictive control (MPC) to incorporate the "worst case" predictive motion of the moving surface. Integrated with an LIP model and walking stability constraints, the CMPC framework generates a set of consistent control inputs considering to anticipated uncertainties of the surface motions. Simulation results and comparison with the regular MPC for bipedal walking are conducted and presented. The results confirm the feasibility and superior performance of the proposed CMPC design over the regular MPC under various motion profiles of moving surfaces.

What problem does this paper attempt to address?

The paper primarily addresses the challenges faced by bipedal robots when walking on dynamically moving surfaces (such as ships or vehicles) and proposes a new control strategy. Specifically, the paper addresses the following key issues: 1. **Control Challenges**: The main challenges faced by bipedal robots when walking on dynamically moving surfaces include: - Limited Balance Control: Due to the inherent instability of bipedal systems, balance control becomes complex. - Unknown Surface Motion: The dynamic characteristics of the moving surface are unknown, which increases the difficulty of control. 2. **Proposed Solution**: To address the above challenges, the authors propose a method called "Contingency Model Predictive Control" (CMPC). This method is based on the Linear Inverted Pendulum Model (LIP) and improves upon the standard Model Predictive Control (MPC) by considering the "worst-case" motion prediction of the moving surface. 3. **Key Contributions**: - **Application of Contingency Concept**: For the first time, the contingency concept is applied to estimate the uncertainty range of the moving surface motion and integrate it into the MPC design to stabilize the walking of the bipedal robot. - **Reduction of Conservatism**: Compared to existing robust MPC methods, the proposed CMPC method reduces excessive conservatism, improves computational efficiency, and ensures performance. 4. **Simulation Validation**: The effectiveness and superiority of the proposed method are validated through simulation experiments. The experimental results show that CMPC can successfully stabilize the robot under different types of surface motion, whereas traditional MPC methods fail to maintain a stable walking posture. In summary, this research provides an innovative and effective control strategy for achieving stable walking of bipedal robots in dynamic environments.

Contingency Model Predictive Control for Bipedal Locomotion on Moving Surfaces with a Linear Inverted Pendulum Model

Nonlinear MPC-Based Control Framework for Quadruped Robots: Touch-Down in Complex Terrain

Robust Locomotion Exploiting Multiple Balance Strategies: An Observer-Based Cascaded Model Predictive Control Approach

Discrete time model predictive control for humanoid walking with step adjustment

Terrain-Adaptive, ALIP-Based Bipedal Locomotion Controller via Model Predictive Control and Virtual Constraints

An MPC-DCM Control Method for a Forward-Bending Biped Robot Based on Force and Moment Control

Nonlinear model predictive control for robust bipedal locomotion: exploring angular momentum and CoM height changes

Model Predictive Control for Motion Planning of Quadrupedal Locomotion.

Chance-Constrained Convex MPC for Robust Quadruped Locomotion Under Parametric and Additive Uncertainties

Global Position Control on Underactuated Bipedal Robots: Step-to-step Dynamics Approximation for Step Planning

Robust-Locomotion-by-Logic: Perturbation-Resilient Bipedal Locomotion via Signal Temporal Logic Guided Model Predictive Control

Real-Time Control of Full Actuated Biped Robot Based on Nonlinear Model Predictive Control

Contingency Model Predictive Control for Linear Time-Varying Systems

Unified Path and Gait Planning for Safe Bipedal Robot Navigation

Legged Robot Locomotion Via Hybrid Zero Dynamics and Model Predictive Control

Balanced Standing on One Foot of Biped Robot Based on Three-Particle Model Predictive Control

Real-Time Footprint Planning and Model Predictive Control Based Method for Stable Biped Walking

Adaptive Model Predictive Control with Data-driven Error Model for Quadrupedal Locomotion

Time-Varying ALIP Model and Robust Foot-Placement Control for Underactuated Bipedal Robot Walking on a Swaying Rigid Surface

Learning Near-global-optimal Strategies for Hybrid Non-convex Model Predictive Control of Single Rigid Body Locomotion

Robust Walking for Humanoid Robot Based on Divergent Component of Motion