Predictive Control for Agile Semi-Autonomous Ground Vehicles using Motion Primitives Real-time control in autonomous systems needs to balance two important issues: model fidelity and computational complexity. Oversimplified models violate the system constraints, whereas using high-fidelity models is computationally demanding. In this paper, the authors try to address the computational expensive nature of using Nonlinear MPC (NMPC) for agile autonomous control in real-time. Due to computational complexity of NMPC, its previous realtime use cases have been limited to low vehicle speeds. Their solution is based on using a hierarchical approach where the high-level Path Planner which uses motion primitives to generate feasible and desired trajectories; and a low level Path Follower is used to follow these paths using a high-fidelity model. Motion primitive idea is based on pre-specified primitives named trims and maneuvers. The trims corresponds to the discrete states in a hybrid system which indicates the dynamics to be followed, whereas maneuvers corresponds control inputs that allows transitioning between these states. This work is novel in using motion primitives for ground autonomous vehicles. Then the authors evaluate their methods on simulation and hardware settings. The methods performs successfully in both cases. An interesting future work is autonomous discovery of trims and maneuvers instead of hardcoding them.