Event
Render-free event simulator accelerates navigation reinforcement learning
1 reports · 1 sources
Dequan Ou and co-authors propose an analytical event simulator that synthesizes event streams directly from scene geometry and ego-motion, without going through image rendering. The authors claim that generating 64x64 event observations on a single GPU takes only 17 microseconds, more than 400 times faster than advanced rendering pipelines; reinforcement learning policies for event-camera obstacle avoidance trained on this converge in under 30 minutes, three orders of magnitude faster than training with rendered events. The authors report that the policy can transfer zero-shot to Isaac Sim ray-traced event simulation; it achieves an 83% success rate on cluttered tracks unseen during training (average target speed 2.4 m/s), and maintains 70% in a texture-dense indoor hall. On three real obstacle tracks of increasing difficulty, without fine-tuning, external processing, or localization, the average success rate is 82% (5/5, 5/6, 4/6 flights).
Report timeline
- Rendering-Free Event Frame Simulation Accelerates Reinforcement Learning for Autonomous Robot Navigation
The authors propose an analytic event simulator that synthesizes event streams directly from scene geometry and ego-motion without image rendering, generating 64x64 event observations in just 17 microseconds on a single GPU, over 400 times faster than state-of-the-art rendering pipelines. Reinforcement learning training for an event-camera obstacle-avoidance policy based on this simulator converges in under 30 minutes, three orders of magnitude faster than training on rendered events, and transfers zero-shot to ray-traced event simulation in Isaac Sim.
arXiv cs.RO ·