REFLEX: A Learned On-Board Reflex for Robust Autonomous Collision Recovery of Soft Quadrotors
We present REFLEX, a collision-recovery flight mode that runs entirely on the flight controller, requiring neither contact sensors nor a companion computer.
ProofPaper ↗
Key points
- Collision-resilient airframes allow quadrotors to survive impacts, but do not ensure autonomous post-collision recovery.
- Innovation gating in stock onboard estimators often rejects valid external velocity measurements after high-speed impacts, compromising recovery.
- Its proprioceptive recurrent policy stabilizes the vehicle without position, velocity, heading, or altitude estimates.
- In hardware experiments, REFLEX recovered 19 of 22 wall impacts at 5.03 $\pm$ 0.09 m/s, compared with 1 of 14 for the matched stock autopilot.
Sources (1)
- [1]REFLEX: A Learned On-Board Reflex for Robust Autonomous Collision Recovery of Soft QuadrotorsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 11, 07:39 PM
We present REFLEX, a collision-recovery flight mode that runs entirely on the flight controller, requiring neither contact sensors nor a companion computer.
Collision-resilient airframes allow quadrotors to survive impacts, but do not ensure autonomous post-collision recovery.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 11, 2026LLM Value Functions for Credit Assignment in Agentic RL
- Oct 10, 2026Physics Foundation Models for Efficient Control of Fluid Flows
- Oct 10, 2026Metric-Driven Reinforcement Learning for Object Hallucination Reduction in LVLM-Based Image Captioning
- Oct 9, 2026Visualizing the World Model in an Agent's Value Function
- Oct 9, 2026Noise Space Learning for VLA Adaptation
- Oct 9, 2026Beyond Temporal Proximity: Explanatory World Models for Credit Assignment in Reinforcement Learning