A Novel Cross-Tree Informed Bidirectional Optimization for Motion Planning of Robots
Year: 2026
Type: article
Abstract: This paper presents a novel cross-tree informed bidirectional RRT* (CTIB-RRT*) algorithm for motion planning of robots, which preserves the rapid feasibility of bidirectional extension and enables effective path optimization. Specifically, the algorithm initially carries out bidirectional tree expansion, where the first successful connection activates rewiring and best parent selection with rapid feasible solution discovery. After each feasible path is obtained, the cross-tree path optimization is enabled to guarantee strict path cost reduction and tree validity. Then, the consecutive state triplets are refined via direct update to reduce the cost without additional candidate collection. When a feasible direct connection is unavailable, hypersphere-based filtering and lazy collision checking are employed to select and evaluate candidate states for computational efficiency, and segment update is performed to exploit the candidate states for path optimization. It is shown that refined solution cost further tightens the informed sampling subset to accelerate convergence of the planner. Simulations in$\mathbb {R}^{2}$,$\mathbb {R}^{6}$and$\mathbb {R}^{12}$tasks, together with standardized MotionBenchMaker benchmarks, are carried out to validate the effectiveness and superiority of the proposed CTIB-RRT* over state-of-the-art planner.
Institution: Southeast University
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $2,645
OpenAlex ID: W7204472121