Abstract
The generation of energy-efficient and dynamic-aware robot motions that satisfy constraints such as joint limits, self-collisions, and collisions with the environment remains a challenge. In this context, Riemannian geometry offers promising solutions by identifying robot motions with geodesics on the so-called configuration space manifold. While this manifold naturally considers the intrinsic robot dynamics, constraints such as joint limits, self-collisions, and collisions with the environment remain overlooked. In this paper, we propose a modification of the Riemannian metric of the configuration space manifold allowing for the generation of robot motions as geodesics that efficiently avoid given regions. We introduce a class of Riemannian metrics based on barrier functions that guarantee strict region avoidance by systematically generating accelerations away from no-go regions in joint and task space. We evaluate the proposed Riemannian metric to generate energy-efficient, dynamic-aware, and collision-free motions of a humanoid robot as geodesics and sequences thereof.
Introduction
Generating robot motion that is simultaneously energy-efficient and safe is a fundamental challenge in robotics. A sophisticated approach involves treating the configuration space as a Riemannian manifold, where the robot's dynamics define a kinetic energy metric. In this geometric framework, the optimal motion corresponds to a geodesic—the shortest path on the manifold. However, standard kinetic energy metrics do not inherently account for obstacles or joint limits, often leading to collisions or constraint violations.
This paper introduces a novel class of region-avoiding metrics that reshape the manifold's geometry around "no-go" zones. By integrating inverse barrier functions directly into the Riemannian metric, the authors ensure that geodesics naturally and strictly avoid prohibited regions.
Methodology
The proposed approach modifies the configuration space manifold through several key innovations:
Inverse Barrier Functions
The authors utilize barrier functions that approach infinity at region boundaries. They demonstrate that inverse barriers ($1/dist$) are superior to traditional exponential barriers. While exponential barriers only provide "soft" avoidance—allowing high-momentum robots to potentially "tunnel" through—inverse barriers create a "hard" geometric wall that geodesics cannot cross.
Direction-Dependent Basis Change
To prevent irregular accelerations in multi-dimensional spaces, the authors implement a change of basis using the Gram-Schmidt process. This ensures that the repelling force generated by the metric always points directly away from the obstacle, independent of the robot's current velocity direction.
Directional Pullbacks
While joint limits are easily defined in joint space, environmental obstacles exist in 3D task space. The authors propose pulling back the direction to the obstacle from task space to joint space using the Jacobian pseudo-inverse. This ensures that the joint-space metric produces physically intuitive repelling accelerations in the real world.
Experimental Evaluation
The approach was evaluated using the KIT Whole-Body Human Motion Database, retargeting human movement sequences (reaching, throwing, pointing, and waving) to an ARMAR-6 humanoid robot. The testing focused on an 8-degree-of-freedom arm, challenging the system with joint limits, self-collisions (e.g., crossing arms), and environmental obstacles (e.g., navigating around a table).
Results and Discussion
The inverse-barrier metric was compared against standard kinetic energy and exponential-barrier metrics.
Safety and Constraint Satisfaction: The proposed metric achieved 0% violations across all tests, including complex "arms crossing" and "waving" tasks. In contrast, baseline methods frequently resulted in significant joint limit violations and self-collisions.
Efficiency: The generated paths remained close to the energy-optimal geodesics of the original kinetic energy metric, deviating only as required to maintain safety. In obstacle avoidance scenarios, such as navigating a table, the inverse-barrier metric successfully guided the robot without getting trapped in local minima.
Conclusion
By embedding safety constraints directly into the geometry of the configuration space, the robot's "natural" behavior becomes inherently safe. This eliminates the need for separate collision-avoidance controllers. Future work will explore velocity-dependent Finsler metrics to distinguish between approaching and receding from obstacles, as well as optimizing computational efficiency for full-body humanoid control.
Summary updated