Reproduce “Gaussian-Process-based Adaptive Tracking Control with Dynamic Active Learning for Autonomous Ground Vehicles” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Kristóf Floch, Tamás Péni, Roland Tóth
An adaptive trajectory tracking control method for autonomous ground vehicles integrates Gaussian Processes (GPs) with dynamic active learning for improved efficiency and robustness. This approach successfully addresses model uncertainty and computational challenges, validated through high-fidelity simulations and real-world F1TENTH experiments.
Key insights
- The combination of LPV model decoupling with sparse Gaussian Process augmentation offers a computationally efficient way to compensate for significant model mismatch in real-time adaptive control.
- Dynamic active learning, which plans trajectories to explicitly explore high-uncertainty regions of the GP model, efficiently gathers informative data to accelerate model refinement in dynamic systems.
- An iterative L2-gain analysis algorithm can provide formal, quantitative performance guarantees ("robust performance certificates") for learning-based control systems under uncertain vehicle dynamics.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.