Research Analyzer
← Back ICRA 2026

Real-Time Learning of Predictive Dynamic Obstacle Models for Robotic Motion Planning

Stella Kombo, Joel Burdick, Masih Haseli, Skylar X. Wei

PDF

AI summary

Key figure (auto-extracted from paper)
An adaptive sliding-window Hankel-DMD framework enables real-time, variance-aware denoising and short-horizon prediction of nonlinear obstacle dynamics from noisy, partial sensor data.
Real-time prediction Hankel-DMD online learning dynamic obstacle modeling denoising robotic motion planning

Problem

Autonomous robots must predict the motions of nearby dynamic agents from partial, noisy, and non-stationary sensor data to ensure safe navigation, but existing methods struggle with real-time adaptation under unknown noise distributions.

Approach

The method continuously embeds streaming measurements into a Hankel matrix, uses a Page matrix with singular value hard thresholding to automatically estimate the system's rank, applies Cadzow projection for denoising, and constructs a time-varying linear predictor for multi-step forecasting.

Key results

  • Proved Page-Hankel rank equivalence for reliable data-driven rank estimation
  • Developed an adaptive rank selection method that automatically tracks non-stationary dynamics
  • Achieved stable, variance-aware denoising and accurate short-horizon predictions under Gaussian and heavy-tailed noise
  • Validated real-time performance and robustness on a dynamic crane testbed

Why it matters

Enables safe, real-time robotic motion planning in dynamic environments by providing reliable, uncertainty-aware predictions from noisy sensor data without offline training.

Abstract

Autonomous systems often must predict the mo- tions of nearby agents from partial and noisy data. This paper asks and answers the question: "can we learn, in real-time, a nonlinear predictive model of another agent’s motions?" Our online framework denoises and forecasts such dynamics using a modified sliding-window Hankel Dynamic Mode Decomposition (Hankel-DMD). Partial noisy measurements are embedded into a Hankel matrix, while an associated Page matrix enables singular-value hard thresholding (SVHT) to denoise the Hankel matrix and estimate its rank. A Cadzow projection enforces structured low-rank consistency, yielding a denoised trajectory and local noise variance estimates. From this representation, a time-varying Hankel-DMD lifted linear predictor is constructed for multi-step forecasts. The residual analysis provides variance- tracking signals that can support downstream estimators and risk-aware planning. We validate the approach in simulation under Gaussian and heavy-tailed noise, and experimentally on a dynamic crane testbed. Results show that the method achieves stable variance-aware denoising and short-horizon prediction suitable for integration into real-time control frameworks.

Index terms

Dynamics Calibration and Identification Machine Learning for Robot Control

Related papers