Generating Diverse Trajectories Using Roadmap Search and Sampling-Based Motion Planning¶
Authors: Evis Plaku, Arben Çela, Erion Plaku
Published: 2026 (Journal Paper)
Source: IEEE Access
DOI: 10.1109/ACCESS.2026.3690566
Summary¶
Abstract¶
This paper presents a motion-planning framework for generating a diverse set of collision-free and dynamically feasible trajectories for robots navigating complex environments. Producing multiple distinct trajectories enables flexible and informed decision-making, offering alternative routes that improve robustness and adaptability in navigation tasks. The framework introduces: (i) a roadmap construction method that captures free-space connectivity and supports candidate path generation, (ii) a roadmap search algorithm based on adaptive costs that produces a pool of spatially distinct paths and selects a subset with low overlap as geometric guides, and (iii) a guided sampling-based motion planner that converts these guides into executable motions while enforcing dynamic feasibility. This approach ensures meaningful spatial variation, converting geometric diversity into distinct dynamically feasible trajectories across different robot models and complex environments. Extensive simulation experiments with car-like and snake-like robots demonstrate that the approach efficiently generates multiple high-quality, spatially diverse and dynamically feasible trajectories in obstacle-dense, challenging environments, outperforming alternative planners.