Balancing experiments on a torque-controlled humanoid with hierarchical inverse dynamics (1305.2042v3)

Abstract: Recently several hierarchical inverse dynamics controllers based on cascades of quadratic programs have been proposed for application on torque controlled robots. They have important theoretical benefits but have never been implemented on a torque controlled robot where model inaccuracies and real-time computation requirements can be problematic. In this contribution we present an experimental evaluation of these algorithms in the context of balance control for a humanoid robot. The presented experiments demonstrate the applicability of the approach under real robot conditions (i.e. model uncertainty, estimation errors, etc). We propose a simplification of the optimization problem that allows us to decrease computation time enough to implement it in a fast torque control loop. We implement a momentum-based balance controller which shows robust performance in face of unknown disturbances, even when the robot is standing on only one foot. In a second experiment, a tracking task is evaluated to demonstrate the performance of the controller with more complicated hierarchies. Our results show that hierarchical inverse dynamics controllers can be used for feedback control of humanoid robots and that momentum-based balance control can be efficiently implemented on a real robot.

• The paper shows that cascading quadratic programs for hierarchical inverse dynamics can robustly balance a torque-controlled humanoid.
• It introduces a constraint simplification that cuts computation time by about 40%, enabling a real-time control loop at 1 kHz.
• The experiments validate a momentum-based controller that effectively handles lateral and sagittal disturbances in dynamic environments.

An Expert Analysis of "Balancing Experiments on a Torque-Controlled Humanoid with Hierarchical Inverse Dynamics"

The paper "Balancing Experiments on a Torque-Controlled Humanoid with Hierarchical Inverse Dynamics" by Herzog et al. presents an in-depth exploration and experimental validation of hierarchical inverse dynamics controllers for torque-controlled humanoid robots. This paper addresses the practical application of these controllers in real-world scenarios characterized by uncertainties such as modeling inaccuracies and sensor noise, a notable challenge for humanoid robotics.

Key Contributions

The research highlights the application of hierarchical inverse dynamics controllers, structured as cascades of quadratic programs (QPs), in achieving robust balance control for humanoid robots. The paper introduces a simplification of dynamic constraints which reduces computational complexity, allowing integration into a real-time control loop operating at 1 kHz. Furthermore, the work experiments with a momentum-based balance controller that adapts to unknown disturbances and evaluates the tracking accuracy for more complex task hierarchies.

Experimental Framework

Herzog and colleagues employed the lower part of the Sarcos Humanoid Robot, equipped with hydraulic actuators and various sensors, to test the proposed control strategies. The experimental setup provided a rigorous environment with real-time constraints, enabling the assessment of the controller's performance in balancing and tracking tasks.

For the balancing experiments, the researchers implemented a momentum-based controller, adapting strategies from prior theoretical advancements in humanoid balance control. The experiments confirmed the controller's efficacy in maintaining balance under lateral and sagittal disturbances, even with additional challenges such as co-planar foot misalignments. These results are significant as they demonstrate the controller's capability to respond effectively to dynamic perturbations, a fundamental requirement for humanoid robots operating in unstructured environments.

In the tracking experiment, a hierarchy of tasks was constructed with physical constraints at the highest priority, followed by center of mass (CoM) tracking tasks and posture maintenance. The experiments demonstrated satisfactory CoM tracking performance, with the controller efficiently managing disturbances while meeting the constraints inherent to the system.

Numerical Results and Implications

The computational efficiency was measured in terms of control loop execution time, where the maximum computation time remained below 1 ms for the tasks performed with 14 degrees of freedom (DoFs). A noteworthy element in the paper was the introduction of constraint simplification, which reduced computation time by approximately 40% in simulations involving all 25 DoFs of the humanoid robot.

These benchmarks establish the hierarchical inverse dynamics approach as a viable method for real-time implementation. The research underscores its potential to enhance feedback control mechanisms for humanoids through reliable handling of model uncertainties and rapid adaptation to unforeseen external forces.

Theoretical and Practical Impacts

On a theoretical level, the integration of momentum control into a hierarchical task prioritization framework offers a structured methodology for handling complex robot dynamics. By maintaining the integrity of task priorities while adapting rapidly to disturbances, this approach aligns well with the objectives of agile, adaptable humanoid systems.

Practically, the demonstrated balancing and tracking capabilities position this method as a promising candidate for applications requiring humanoid robots to perform in dynamic or unpredictable environments, such as disaster response scenarios. The success of these experiments also suggests possibilities for extending the framework to encompass more sophisticated multi-tasking and interaction capabilities in humanoids.

Future Directions

Future research could delve into the optimization of the QP solver to handle more complex task spaces and greater numbers of DoFs, extending the application of this method to full humanoid bodies and potentially multi-contact dynamics. Investigating the integration of advanced dynamics models with enhanced state estimation techniques could further improve the balance and agility of humanoid robots.

In conclusion, this paper contributes significantly to the field by bridging theoretical advancements in hierarchical control strategies with practical, real-world application, marking a step forward in the development of versatile, autonomous humanoid robots.