Momentum Control with Hierarchical Inverse Dynamics on a Torque-Controlled Humanoid (1410.7284v2)

Abstract: Hierarchical inverse dynamics based on cascades of quadratic programs have been proposed for the control of legged robots. They have important benefits but to the best of our knowledge have never been implemented on a torque controlled humanoid where model inaccuracies, sensor noise and real-time computation requirements can be problematic. Using a reformulation of existing algorithms, we propose a simplification of the problem that allows to achieve real-time control. Momentum-based control is integrated in the task hierarchy and a LQR design approach is used to compute the desired associated closed-loop behavior and improve performance. Extensive experiments on various balancing and tracking tasks show very robust performance in the face of unknown disturbances, even when the humanoid is standing on one foot. Our results demonstrate that hierarchical inverse dynamics together with momentum control can be efficiently used for feedback control under real robot conditions.

• The paper reformulates hierarchical inverse dynamics using LQR to achieve efficient real-time momentum control in torque-controlled humanoids.
• It demonstrates robust stabilization even under significant external forces, with experimental tests showing peak forces up to 290 N.
• The study improves computational efficiency by 40% through a streamlined dynamic formulation, supporting a 1kHz control loop in dynamic environments.

Hierarchical Inverse Dynamics for Humanoid Momentum Control

The paper under discussion presents an experimental validation of hierarchical inverse dynamics on a torque-controlled humanoid robot, focusing on momentum control. It ambitiously tackles the challenges of real-time implementation in environments with model inaccuracies and sensor noise. The authors propose a streamlined problem formulation using a hierarchy of quadratic programs (QPs), which are adept at managing complex multi-contact tasks.

The main contribution lies in reformulating existing hierarchical inverse dynamics algorithms to achieve efficient real-time feedback control on a humanoid robot operating in non-ideal conditions. The proposed approach integrates momentum-based control within a task hierarchy and employs Linear Quadratic Regulator (LQR) design to translate desired closed-loop behavior into optimal feedback gains for linear and angular momentum regulation. This hybrid control strategy is evaluated through rigorous experimental tests, demonstrating robust performance across various balance and tracking tasks, including when the robot is standing on one foot.

Key experimental results suggest that the use of hierarchical inverse dynamics in conjunction with LQR momentum control leads to significant improvements in feedback control efficacy. The humanoid maintained stability under substantial external disturbances, with peak forces reaching up to 290 N and impulses up to 9.5 Ns. These quantitative results highlight the approach's competence in maintaining the center of gravity (CoG) tracking and damping changes in momentum, underscoring its potential application in dynamic and highly interactive scenarios.

The proposed method also emphasizes computational efficiency. The reformulated solver structure, incorporating a simplification of the dynamic equations, facilitates a reduction in computation time by 40% compared to traditional approaches. This efficiency is crucial for maintaining a 1kHz control loop, necessary for real-world robotic applications.

This work has broad implications for both theoretical advancements and practical robotics applications. Theoretically, it underscores the viability of using hierarchical task priorities in conjunction with precise momentum control to achieve robust humanoid robot performance. Practically, the paper points to the potential of deploying humanoid robots in dynamic environments, such as disaster relief scenarios, where both stability and adaptability to disturbances are critical.

Moving forward, the integration of high-level planners to enable more complex task management, such as bipedal walking, and the refinement of dynamic models through parameter identification could enhance the robustness and adaptability of such systems. The momentum-based hierarchical control paradigm has demonstrated promise for advancing autonomous humanoid capabilities, paving the way for further exploration into agile and interactive robotic systems.