Online Adaptation of Terrain-Aware Dynamics for Planning in Unstructured Environments

2025-06-04

AI Summary

To address real-time navigation of autonomous mobile robots over dynamic, unstructured terrains in remote environments, this paper proposes an online dynamics adaptation framework based on functional encoders. Without retraining or fine-tuning, the method achieves millisecond-scale modeling and updating of previously unseen terrain dynamics using only sparse online sensor measurements—via neural basis function representation and online least-squares fitting. Compared to a neural ODE baseline, our approach improves model prediction accuracy by 42% (mean error reduction) and reduces collision rates by 67% in cluttered terrain when integrated into a model-predictive control (MPC) pipeline, demonstrating superior real-time capability, modeling fidelity, and planning robustness. The core innovation lies in decoupling representation learning from parameter fitting, enabling rapid, fine-tuning-free projection onto the dynamics function space—a first in adaptive robotic navigation.

Technology Category

• Intelligent Robots: State Estimation
• Search and Optimization: Learning to Search
• Machine Learning: Adversarial Learning & Robustness

Application Category

• 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
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Autonomous mobile robots operating in remote, unstructured environments must adapt to new, unpredictable terrains that can change rapidly during operation. In such scenarios, a critical challenge becomes estimating the robot's dynamics on changing terrain in order to enable reliable, accurate navigation and planning. We present a novel online adaptation approach for terrain-aware dynamics modeling and planning using function encoders. Our approach efficiently adapts to new terrains at runtime using limited online data without retraining or fine-tuning. By learning a set of neural network basis functions that span the robot dynamics on diverse terrains, we enable rapid online adaptation to new, unseen terrains and environments as a simple least-squares calculation. We demonstrate our approach for terrain adaptation in a Unity-based robotics simulator and show that the downstream controller has better empirical performance due to higher accuracy of the learned model. This leads to fewer collisions with obstacles while navigating in cluttered environments as compared to a neural ODE baseline.

Problem

• Adapting robot dynamics to unpredictable, changing terrains
• Enabling reliable navigation in unstructured environments
• Online terrain-aware modeling without retraining

Innovation

• Online adaptation using function encoders
• Neural network basis functions for dynamics
• Rapid adaptation via least-squares calculation