Abstract
In recent years the problem of planning paths through complex obstacles while respecting kinematic constraints has seen increased attention. Many have found that respecting curvature constraints is particularly difficult without substan-tially slowing convergence speed. Batch Informed Trees (BIT*) is a path-planning algorithm that has been shown to converge rapidly in large environments without considering kinematic constraints. This work proposes an extension to BIT* that employs fillets as motion primitives, enabling the incorporation of curvature constraints into the planning process. Path-length heuristics for fillet-based planning are introduced to accelerate convergence. Comparisons to pre-existing approaches are made with an Unmanned Aerial Vehicle (UAV) simulation modeled off of Manhattan, New York.