Uncertainty Guided Exploratory Trajectory Optimization for Sampling-Based Model Predictive Control
Oguzhan Goktug Poyrazoglu, Yukang Cao, Rahul Moorthy Mahesh, Volkan Isler
AI summary
Problem
Sampling-based trajectory optimization methods struggle with poor initialization and limited exploration, frequently converging to local minima in complex environments because they only perturb actions without accounting for system dynamics.
Approach
The method models trajectories as probability distributions using uncertainty ellipsoids and enforces diversity by maximizing the Hellinger distance between distributions, enabling systematic configuration-space exploration before local MPC refinement.
Key results
- Introduces UGE-TO using uncertainty ellipsoids and Hellinger distance for distributional separation
- Integrates UGE-TO into sampling-based MPC to improve nominal trajectory selection
- Achieves 72.1% faster convergence in obstacle-free and 66% faster with 6.7% higher success rate in cluttered environments
- Validated across diverse simulations and real-world navigation experiments
Why it matters
Provides a robust, dynamics-aware exploration strategy for sampling-based MPC, benefiting robotic navigation and manipulation in complex, multi-modal environments.
Abstract
Trajectory optimization depends heavily on ini- tialization. In particular, sampling-based approaches are highly sensitive to initial solutions, and limited exploration frequently leads them to converge to local minima in complex environ- ments. We present Uncertainty Guided Exploratory Trajectory Optimization (UGE-TO), a trajectory optimization algorithm that generates well-separated samples to achieve a better cover- age of the configuration space. UGE-TO represents trajectories as probability distributions induced by uncertainty ellipsoids. Unlike sampling-based approaches that explore only in the action space, this representation captures the effects of both system dynamics and action selection. By incorporating the impact of dynamics, in addition to the action space, into our distributions, our method enhances trajectory diversity by enforcing distributional separation via the Hellinger distance between them. It enables a systematic exploration of the con- figuration space and improves robustness against local minima. Further, we present UGE-MPC, which integrates UGE-TO into sampling-based model predictive controller methods. Experi- ments demonstrate that UGE-MPC achieves higher exploration and faster convergence in trajectory optimization compared to baselines under the same sampling budget, achieving 72.1% faster convergence in obstacle-free environments and 66% faster convergence with a 6.7% higher success rate in the cluttered environment compared to the best-performing base- line. Additionally, we validate the approach through a range of simulation scenarios and real-world experiments. Our results indicate that UGE-MPC has higher success rates and faster convergence, especially in environments that demand significant deviations from nominal trajectories to avoid failures. The project and code are available at https://ogpoyrazoglu. github.io/cuniform_sampling/.