Stability-Aware Retargeting for Humanoid Multi-Contact Teleoperation

TL;DR

This work tackles stability during humanoid teleoperation in multi-contact tasks where hand contacts push limits of actuation and friction. It introduces a centroidal stability region–based retargeting method, with an efficient LP-sensitivity gradient that informs local adjustments to contact positions and posture to increase the stability margin $m$. The approach uses a shared-control teleoperation framework, applying contact retargeting when gradient-based previews indicate improvement and posture retargeting when sensitivity is high, and it is validated with both simulation and hardware, showing improved impulse resilience and joint-torque margins. Overall, the method demonstrates real-time stability augmentation for contact-rich humanoid manipulation, enabling more robust teleoperation in challenging environments.

Abstract

Teleoperation is a powerful method to generate reference motions and enable humanoid robots to perform a broad range of tasks. However, teleoperation becomes challenging when using hand contacts and non-coplanar surfaces, often leading to motor torque saturation or loss of stability through slipping. We propose a centroidal stability-based retargeting method that dynamically adjusts contact points and posture during teleoperation to enhance stability in these difficult scenarios. Central to our approach is an efficient analytical calculation of the stability margin gradient. This gradient is used to identify scenarios for which stability is highly sensitive to teleoperation setpoints and inform the local adjustment of these setpoints. We validate the framework in simulation and hardware by teleoperating manipulation tasks on a humanoid, demonstrating increased stability margins. We also demonstrate empirically that higher stability margins correlate with improved impulse resilience and joint torque margin.

Figures (11)

• Figure 1: Robot performing a teleoperated manipulation task, in which the left hand contacts the wall to extend the reachable workspace of the robot and retrieve a canister. The robot's stability margin is improved by locally adjusting the posture and contact setpoints commanded from the operator.
• Figure 2: Depiction of the stability region. Vertex $i$ corresponds to a direction $\mathbf{a}_i$ and is computed by Eq. (\ref{['eq:CoMLP']}). The stability margin $m$ is the minimum distance from the CoM $\mathbf{c}$ to the region boundary.
• Figure 3: The 2d stability region encodes both actuation and friction constraints. When actuation constrained, the boundary of the stability region corresponds to the boundary of a force polytope. $\nabla m(\mathbf{q})$ is therefore a whole-body motion which rotates the force polytope and allows for increased contact force.
• Figure 4: Control flow for contact and posture retargeting. The contacting hand is adjusted prior to loading the hand by previewing the stability region at the robot's current configuration. The area gradient is used to grow the preview region. When the posture sensitivity$s_q$ is sufficiently high, the stability margin gradient informs the adjustment of the posture setpoints.
• Figure 5: The stability region is incrementally updated using a set of fixed query directions, in which one full Linear Program solve (Eq. (\ref{['eq:CoMLP']})) and two fast, fixed-basis solves (Eq. (\ref{['eq:fixedBasisUpdate']})) are performed.