Event
Differentiable Simulation Trains Quadrotor to Fly Through Dynamic Narrow Gaps
1 reports · 1 sources
The research team proposed a differentiable modeling method for moving narrow gaps, transforming the sparse event of a quadrotor flying through a dynamic narrow gap into a dense path objective, and on this basis trained, from random initialization in a single stage, a policy that outputs commands directly from visual features, without a teacher, planner, or distillation. The team reports that in simulation the path gradient is clearly superior to the score-function gradient, and that after a few minutes of training on a single GPU the success rate reached 99.3%. The policy was then transferred directly to a real quadrotor, relying only on onboard vision and inertial measurements, and successfully flew through the dynamic narrow gap 8 times out of 10 physical trials.
Report timeline
- Research team uses differentiable simulation to learn quadrotor dynamic narrow-gap traversal from scratch
The research proposes a differentiable moving narrow-gap modeling method that gates centering, alignment, and smooth collision terms by signed distance to a time-varying gap plane, turning sparse traversal events into dense pathwise objectives, and integrates this into a visual-feature-to-command policy trained in a single stage from random initialization, without a teacher, planner, or distillation. In simulation, path gradients clearly outperform score-function gradients, reaching a 99.3% success rate after a few minutes of training on a single GPU. The policy transfers directly to a real quadrotor using only onboard vision and inertial measurements, successfully traversing the dynamic narrow gap in 8 of 10 real-world trials.
arXiv cs.RO ·