Problem
Research questions and friction points this paper is trying to address.
Introduces VectorTSP with acceleration constraints
Combines city order and physical trajectory planning
Demonstrates NP-hardness and suboptimal naive approaches
VectorTSP: A Traveling Salesperson Problem with Racetrack-like acceleration constraints
🤖 AI Summary
This paper studies the Vector Traveling Salesman Problem (VectorTSP), a variant incorporating vector-valued motion constraints derived from the Racetrack game: position updates are governed by discrete velocity and acceleration dynamics, where velocity is unbounded but state evolution is coupled to the current velocity. The problem jointly optimizes city visitation order and physically feasible trajectories—yielding a tightly coupled bilevel planning formulation. Our key contribution is the rigorous integration of continuous kinematic constraints—including bounded acceleration and minimum turning radius—into the TSP modeling framework, enabling exact and efficient heuristic solutions in the vector state space. We combine state lattices, A* search, dynamic programming, and geometric pruning to ensure kinematically feasible trajectory generation and cost propagation. Evaluated across UAV and autonomous vehicle scenarios, our algorithm achieves millisecond-scale computation with <3% suboptimality, significantly outperforming conventional TSP solvers.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Discrete acceleration constraints
Interactive protocol algorithm
A* adapted trajectory oracle
🔎 Similar Papers
No similar papers found.
