Designing a Hardware-Agnostic Interface Between Route and Trajectory Planning in Self-Driving Cars

Software reusability is critical to the rapid development and certification of autonomous vehicles (AVs). However, little attention has been given to designing AV path planners for increased software reusability. In this paper we design a trajectory-optimizing nonlinear MPC planner that takes a driveable corridor, desired speed profile and constraints as input to compute the AV's trajectory. These inputs are hardware-independent, confining hardware dependencies to the planner software itself and clearly separating higher-level route planning from hardware-dependent algorithms. We implement the planner in simulation to demonstrate its feasibility in representing common traffic scenarios.