Reproduce “Stochastic Multi-Objective Kinodynamic Planning Against Adversaries” 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
Thomas Marshall Vielmetti, Daniel Cherenson, Dimitra Panagou
Researchers at the University of Michigan developed a framework for multi-objective kinodynamic planning that balances execution cost and safety risk in stochastic, adversarial environments. Their Stochastic Multi-Objective Stable Sparse RRT (SMO-SST) algorithm consistently generated superior Pareto fronts, achieving lower minimum risk and cost while accurately accounting for reactive adversaries through closed-loop policy evaluation.
Key insights
- Planning in the space of closed-loop policies, rather than open-loop trajectories, enables more accurate and less conservative risk assessment against reactive adversaries.
- Directly integrating sample-based Monte Carlo rollouts for risk evaluation is effective for capturing complex, non-Gaussian interactions with stochastic hybrid adversaries.
- Sparse search tree structures with multi-objective dominance-based pruning (SMO-SST) can significantly improve the quality and diversity of the discovered Pareto front in multi-objective planning problems.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.