Reproduce “NeuroHJR: Hamilton-Jacobi Reachability-based Obstacle Avoidance in Complex Environments with Physics-Informed Neural Networks” 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
Granthik Halder, Rudrashis Majumder, Rakshith M R, Rahi Shah, Suresh Sundaram
NeuroHJR presents a Physics-Informed Neural Network (PINN) framework that enables real-time, Hamilton-Jacobi Reachability (HJR)-based obstacle avoidance for autonomous ground vehicles. This approach achieved an average 24.6% reduction in travel time and a 2.91% reduction in path length compared to conventional HJR methods, effectively addressing the scalability limitations of HJR while preserving safety.
Key insights
- Physics-Informed Neural Networks (PINNs) can effectively approximate complex Hamilton-Jacobi Reachability (HJR) solutions, alleviating the computational burden and "curse of dimensionality" associated with traditional grid-based methods.
- Integrating the HJB Partial Differential Equation directly into the neural network's loss function enables learning physically consistent and formally safe control policies.
- A larger sensor detection range significantly improves both travel time and path efficiency by allowing earlier detection and more optimal evasive maneuver planning.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.