I am excited to share that our paper, “Exploring How Non-Prehensile Manipulation Expands Capability in Robots Experiencing Multi-Joint Failure”, has been accepted for presentation at IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) 2024!

Paper Summary

In this work, we explore how non-prehensile manipulation (NPM) and whole-body interaction can enable robotic manipulators to retain functionality despite experiencing locked multi-joint (LMJ) failures. LMJ failures—where two or more joints become inoperable—severely restrict a robot’s motion and task execution capabilities. Our approach introduces:

  1. Failure-Constrained Workspace Modeling – Defining the reachability and action space of a failed manipulator.
  2. Kinodynamic Mapping of Non-Prehensile Actions – Precomputing viable NPM actions such as pushing and poking to maintain task success.
  3. A Simulation-Driven Manipulation Planner – Leveraging a sim-in-the-loop approach to select the best non-prehensile actions dynamically.

Our experimental evaluation demonstrates that this method can expand a robot’s failure-constrained workspace by up to 79%, and increase task success rates to 88.9% even when the end-effector is unusable.

Impact & Future Directions

This research is a step toward making fault-tolerant robotic manipulation a reality, particularly in high-stakes applications such as space robotics, remote surgery, and industrial automation. By shifting manipulation paradigms beyond traditional grasping, we open new possibilities for robots to autonomously adapt to unexpected failures.

We look forward to presenting this work at IROS 2024 and engaging with the robotics community on the future of failure-resilient manipulation systems. If you’re attending the conference, come check out our talk!