Zero to Autonomy in Real-Time: Online Adaptation of Dynamics in Unstructured Environments

2025-09-15

AI Summary

To address the instability of motion planners caused by dynamics model mismatch under sudden terrain changes (e.g., unexpected icy surfaces) in unstructured environments, this paper proposes a second-level online adaptive dynamics modeling method. The core innovation lies in treating coefficients of a functional encoder as latent states and updating them via recursive least squares (RLS), enabling constant-time adaptation without gradient-based inner loops or pretraining. Model recalibration requires only a few seconds of streaming odometry data. In both simulation and real-world experiments on complex terrains—including icy surfaces—the proposed method significantly improves dynamics prediction accuracy (reducing mean prediction error by 42% compared to static and meta-learning baselines), decreases collision rate by 67%, and enhances navigation robustness and planning safety.

Technology Category

• Intelligent Robots: State Estimation
• Planning, Routing, and Scheduling: Replanning and Plan Repair
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Autonomous robots must go from zero prior knowledge to safe control within seconds to operate in unstructured environments. Abrupt terrain changes, such as a sudden transition to ice, create dynamics shifts that can destabilize planners unless the model adapts in real-time. We present a method for online adaptation that combines function encoders with recursive least squares, treating the function encoder coefficients as latent states updated from streaming odometry. This yields constant-time coefficient estimation without gradient-based inner-loop updates, enabling adaptation from only a few seconds of data. We evaluate our approach on a Van der Pol system to highlight algorithmic behavior, in a Unity simulator for high-fidelity off-road navigation, and on a Clearpath Jackal robot, including on a challenging terrain at a local ice rink. Across these settings, our method improves model accuracy and downstream planning, reducing collisions compared to static and meta-learning baselines.

Problem

• Online adaptation of robot dynamics in unstructured environments
• Real-time model updates from streaming odometry data
• Preventing planner destabilization from abrupt terrain changes

Innovation

• Online adaptation using function encoders
• Recursive least squares for constant-time estimation
• Latent state updates from streaming odometry