Mixture of Flow Matching and Particle Filter for Proprioception-Based Online Contact Estimation

Key points

• We consider a robot manipulator equipped with joint torque sensors and a base force/torque sensor; such a proprioceptive-only setup presents a challenging inverse problem characterized by (i) multi-modality from non-unique solutions, (ii) temporal dependencies in state evolution, and (iii) discrete contact events involving contact addition/removal.1
• Existing approaches, such as particle filters (PFs) and learning-based methods, often struggle to address these challenges simultaneously.1
• To overcome this, ContactFlowFilter integrates a generative model into the PF framework.1
• The recursive structure of the PF captures temporal dependencies, while the flow matching model serves as an auxiliary proposal distribution to track multi-modal beliefs and detect contact events.1

Sources (1)

• [1]Mixture of Flow Matching and Particle Filter for Proprioception-Based Online Contact Estimation
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 21, 11:00 AM

  In this article, we propose ContactFlowFilter, a novel approach for contact estimation that determines the number of active contacts, their locations, and the associated force vectors.

  We consider a robot manipulator equipped with joint torque sensors and a base force/torque sensor; such a proprioceptive-only setup presents a challenging inverse problem characterized by (i) multi-modality from non-unique solutions, (ii) temporal dependencies in state evolution, and (iii) discrete contact events involving contact addition/removal.

Extractive summary: sentences quoted from the sources.