Multimodal Interaction Estimation for Contact-Aware Adaptive Impedance Control

Key points

• With advancements in learning-based policies, robots are increasingly being deployed in human-centric environments, such as manufacturing floors and homes, where safe and precise operation requires compensating for unknown loads while remaining compliant during human interactions.1
• Reliably distinguishing between these interaction types is challenging with a single sensing modality, as force signals cannot separate payloads from human pushes, and vision is susceptible to occlusions.1
• The classifier drives a system identification framework that selectively compensates for unknown payloads and a variable-impedance controller that switches between adaptive and reactive modes, while simultaneously estimating interaction forces for downstream tasks, such as human intent estimation.1
• We validate the framework under external perturbations and unknown loads across single-arm manipulation, deployment with an imitation-learning-based policy, intent-driven target switching, and zero-shot bimanual deployment on two Franka arms lifting a box of unknown mass, demonstrating generalization to closed kinematic chains and unseen setups, enabling safe and precise deployment.1

Sources (1)

• [1]Multimodal Interaction Estimation for Contact-Aware Adaptive Impedance Control
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 16, 09:00 PM

  We propose a task-agnostic interaction estimation and adaptive control framework that addresses this challenge by fusing force history and 3D point cloud through a transformer-based classifier.

  With advancements in learning-based policies, robots are increasingly being deployed in human-centric environments, such as manufacturing floors and homes, where safe and precise operation requires compensating for unknown loads while remaining compliant during human interactions.

Extractive summary: sentences quoted from the sources.