Search-Based Optimal Motion Planning for Automated Driving

Abstract

This paper presents preliminary work on learning the search heuristic for the optimal motion planning for automated driving in urban traffic. Previous work considered search-based optimal motion planning framework (SBOMP) that utilized numerical or model-based heuristics that did not consider dynamic obstacles. Optimal solution was still guaranteed since dynamic obstacles can only increase the cost. However, significant variations in the search efficiency are observed depending whether dynamic obstacles are present or not. This paper introduces machine learning (ML) based heuristic that takes into account dynamic obstacles, thus adding to the performance consistency for achieving real-time implementation.

Fail-Safe Motion Planning for Online Verification of Autonomous Vehicles Using Convex Optimization

“…Graph-search approaches, such as state lattices, are yet another form of discrete planning techniques and work on fixed graph structures [24]- [29]. They obtain sets of trajectories whose goal states are vertices in a fixed predefined grid, resulting in a lattice structure.…”

Safe Trajectory Generation for Complex Urban Environments Using Spatio-Temporal Semantic Corridor

“…1) Desired state constraints: First of all, the generated trajectory should start from the given initial state [σ (0) t0 , σ (1) t0 , σ (2) t0 ] and terminate at the given goal state [σ (0) tn , σ (1) tn , σ (2) tn ] for σ ∈ {s, l}, where σ (k) t denotes the k-th-order derivative at time t. Specifically, this requires enforcing equality constraints for the first and last segment as follows,…”

Jointly Learnable Behavior and Trajectory Planning for Self-Driving Vehicles

“…Gu et al [4] propose a multi-phase decision making framework where a traffic-free reference planning over a long range is performed first, followed by a traffic-based refinement where other actors are taken into account; lastly, a final step of local trajectory planning addresses the short-term motion horizon in a refined manner. In [1], a search-based behavior planning approach that is capable of handling hundreds of variants in real-time was proposed, addressing the limitation of [3]. This is achieved by representing the search space and the driving constraints with a geometric representation that is amenable to modeling predictive control schemes, and using an explicit cost-to-go map.…”