Floating-Base Humanoid Robot Motion Tracking with Contact Force Constraints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing control methods for humanoid robots fail to effectively manage the dynamics of motion capture data, leading to instability and inability to balance during complex tasks, especially when interacting with uneven environments, as they primarily focus on kinematics rather than dynamics and do not account for strict contact force constraints.
Innovation Solution
A controller with two main components: a proportional-derivative (PD) controller to compute desired accelerations and a joint torque optimization module that computes optimal joint torques and contact forces, decoupling the optimization into simpler sub-problems to ensure real-time feasibility under contact force constraints, allowing the robot to track human motion capture data while maintaining balance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human motion tracking control is implemented for floating-base humanoid robots, then motion reproduction accuracy is improved, but contact force constraints are violated leading to instability
Solution Approach 1:
The control problem is segmented into two separate optimization problems: (1) computing desired accelerations from motion capture data, and (2) computing joint torques and contact forces to achieve those accelerations while satisfying constraints. This segmentation allows independent optimization of tracking accuracy and constraint satisfaction.
Solution Approach 2:
The controller transforms the motion capture data into desired accelerations, then solves for contact forces and joint torques that satisfy both the desired motion and the contact force constraints. By changing the parameter representation from direct position control to acceleration-based control with constraint satisfaction, the system achieves both accuracy and reliability.
2Device complexity
If simplified balance control is used, then computational complexity is reduced, but tracking accuracy and stability deteriorate
Solution Approach 1:
The balance control is extracted and integrated into the main control loop rather than being a separate simplified component. The controller computes desired accelerations that inherently account for balance requirements, and the contact force optimization ensures stability is maintained throughout the motion execution.
3Stability of the object's composition
If full dynamics consideration is implemented, then motion stability is improved, but computational complexity increases
Solution Approach 1:
The complex dynamics problem is segmented into manageable sub-problems: computing desired accelerations from motion capture data, then solving a constrained optimization problem for contact forces and joint torques. This segmentation makes the full dynamics consideration computationally tractable while maintaining stability.
Solution Approach 2:
The controller computes full dynamics considering all constraints, but applies them partially through the optimization framework that balances tracking accuracy with constraint satisfaction. The contact force constraints are enforced exactly, while the motion tracking is optimized within those constraints.
Data Source
AI summary
A controller for floating-base humanoid robots that can track motion capture data while maintaining balance. Briefly, the controller includes a proportional-derivative (PD) controller that is adapted to compute the desired acceleration to track a given reference trajectory at every degree-of-freedom (DOF) of the robot including the six unactuated ones of the floating base. Second, the controller includes a component (joint torque optimization module) that computes the optimal joint torques and contact forces to realize the desired accelerations given by the first component (i.e., the PD controller). The joint torque optimization module performs this computation considering the full-body dynamics of the robot and the constraints on contact forces. The desired accelerations may not be feasible for the robot due to limits in normal contact forces and friction (e.g., the robot sometimes cannot exactly copy or perform the modeled human motion defined by motion capture data).