BRACE: Belief-guided Residual Adaptive Control for Safe Navigation under Unknown Disturbances
Organizations: Reliable Autonomous Systems Lab, Massachusetts Institute of Technology, Cambridge, MA 02139, USA · Automatic Control Laboratory (IfA), ETH Zürich, 8092 Zürich, Switzerland
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 Residual Adaptive Control (BRACE), a hierarchical reinforcement learning (RL) framework for reach-avoid navigation under unknown, bounded, spatially varying disturbances. BRACE infers a disturbance belief from model prediction errors and uses that to condition a high-level policy that adapts the reference trajectory. The policy combines a nominal and a belief-guided residual term, trained jointly with a likelihood-ratio 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.
Figures & tables
| Module | Parameter | Value |
| IMM | Process noise | , |
| Measurement | ||
| Transitions | (slow/fast order) | |
| Initialization | Means ; covariances ; mode probabilities | |
| LQR | Horizon | s (20 steps) |
| Weights | ; |
| Quadrant | Corridor | |||||||
| Method | SR | SR | SR | SR | ||||
| Centerline | 7.7 | 5.5 | 0.0 | 0.0 | ||||
| DA-MPC | 40.2 | 46.7 | 45.7 | 65.2 | ||||
| RP-MPC | 43.9 | 64.6 | 45.5 | 76.5 | ||||
| BRACE-nominal-only | 14.7 | 15.9 | 0.0 | 0.0 | ||||
| Quadrant | Corridor | |||
|---|---|---|---|---|
| Method | ||||
| BRACE-nominal-only | 14.7–43.2 | 15.9–52.0 | 0.0–16.7 | 0.0–20.1 |
| No-belief | 35.7–44.6 | 49.8–60.1 | 43.9–49.5 | 62.7–85.9 |
| No-residual | 48.5–56.3 | 70.5–78.7 | 50.8–60.3 | 83.2–88.3 |
| No-reweight | 41.8–53.8 | 42.1–65.8 | 43.9–58.1 | 67.5–95.0 |
| Frozen-nominal | 50.1–53.8 | 61.7–68.9 | 53.5–62.1 | 84.6–94.2 |
Adaptive Control
Adaptive control lets systems update models, policies, or controller parameters online to maintain performance as dynamics and disturbances change.
See all 157 papers