BRACE: Belief-guided Residual Adaptive Control for Safe Navigation under Unknown Disturbances
BRACE: Belief-guided Residual Adaptive Control for Safe Navigation under Unknown Disturbances
This research presents BRACE, a hierarchical reinforcement learning framework designed for reach-avoid navigation under unknown, bounded, and spatially varying disturbances. The system infers disturbance beliefs from model prediction errors to condition a high-level policy that adapts reference trajectories by combining nominal and belief-guided residual terms. Evaluated across three simulated quadcopter environments with abruptly switching wind fields and validated on Crazyflie hardware with fan-generated disturbances, BRACE demonstrates improved reach-avoid performance compared to model predictive control and reinforcement learning baselines.
The paper introduces BRACE, a hierarchical reinforcement learning framework for safe navigation under unknown disturbances, evaluated in simulated quadcopter environments and on Crazyflie hardware with fan-generated disturbances.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- Environment
- SimulationOpen source S4
- Control
- Reported autonomousOpen source S2
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
io advantage reweighting heuristic. We evaluate BRACE in three simulated quadcopter environments with abruptly switching wind fields, and validate it on Crazyflie hardware with fan-generated disturbances. Across these settings, BRACE improves reach-avoid performance over model predictive control and RL baselines.
Open source S4
arXiv:2611.10938v1 Announce Type: new Abstract: Disturbances encountered during autonomous navigation generally vary across space, often discontinuously, as obstacles redirect wind and changing terrain alters traction. Tracking controllers typically assume sufficient control authority to mitigate disturbances and maintain safety; the regime where this assumption fails remains underexplored. We present Belief-guided R
Open source S2
Source：arXiv Robotics — research abstracts · arxiv.org