Toward Optimal Multi-Vehicle Path Planning With Hybrid Roadmaps

Effective path planning for multi-vehicle autonomous systems is essential to guarantee safe and efficient operation in industrial environments composed of both structured and unstructured settings. In the proposed framework, the overall environment is explicitly modeled as the integration of static and dynamic areas, each exhibiting distinct spatial and temporal properties. Static areas correspond to structured zones, such as corridors or production lines, and are represented by roadmaps with fixed topology. Dynamic areas, instead, represent unstructured portions of the environment and are modeled as gridmaps to enable adaptive navigation under changing occupancy conditions. Building upon this representation, a hybrid roadmap unifying static and dynamic areas is proposed, with a hierarchical control architecture for coordinated multi-vehicle navigation. The high-layer planner determines the optimal sequence of areas to traverse by solving a multi-objective optimization (MOO) problem that balances efficiency, safety, and traffic-related factors. On the other hand, the low-layer planner generates feasible paths within each area according to its structural constraints. The MOO formulation is embedded in a bi-level optimization framework, where the outer-level employs a quasi-adaptive self-tuning process for objective weights, addressing the limitations of subjective or suboptimal weight selection. Experimental validation across two realistic industrial layouts demonstrates improved mission efficiency, reduced travel-time variability, and scalability with fleet size, confirming the effectiveness of the proposed method for Industry 4.0 applications.