# Bi-Level Route Optimization and Path Planning with Hazard Exploration

arXiv: https://arxiv.org/abs/2503.24044 · HTML: https://mlsanity.com/paper/2503.24044/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 31 Mar 2025.

## Built on by (most cited first, 1 of 1)

- [Bayesian Ambiguity Contraction-based Adaptive Robust Markov Decision Processes for Adversarial Surveillance Missions](https://mlsanity.com/paper/2512.01660/) (2025)

## Abstract (arXiv)

Effective risk monitoring in dynamic environments such as disaster zones requires an adaptive exploration strategy to detect hidden threats. We propose a bi-level unmanned aerial vehicle (UAV) monitoring strategy that efficiently integrates high-level route optimization with low-level path planning for known and unknown hazards. At the high level, we formulate the route optimization as a vehicle routing problem (VRP) to determine the optimal sequence for visiting known hazard locations. To strategically incorporate exploration efficiency, we introduce an edge-based centroidal Voronoi tessellation (CVT), which refines baseline routes using pseudo-nodes and allocates path budgets based on the UAV's battery capacity using a line segment Voronoi diagram. At the low level, path planning maximizes information gain within the allocated path budget by generating kinematically feasible B-spline trajectories. Bayesian inference is applied to dynamically update hazard probabilities, enabling the UAVs to prioritize unexplored regions. Simulation results demonstrate that edge-based CVT improves spatial coverage and route uniformity compared to the node-based method. Additionally, our optimized path planning consistently outperforms baselines in hazard discovery rates across a diverse set of scenarios.