Parametrized Maneuvers Governor for Decision Making in Automated Driving

    •  Di Cairano, S., Skibik, T., Vinod, A.P., Weiss, A., Berntorp, K., Okura, Y., "Parametrized Maneuvers Governor for Decision Making in Automated Driving" in Nonlinear and Constrained Control - Applications, Synergies, Challenges and Opportunities., June 2024.
      BibTeX TR2024-086 PDF
      • @incollection{DiCairano2024jun,
      • author = {Di Cairano, Stefano and Skibik, Terrence and Vinod, Abraham P. and Weiss, Avishai and Berntorp, Karl and Okura, Yuichi}},
      • title = {Parametrized Maneuvers Governor for Decision Making in Automated Driving},
      • booktitle = {Nonlinear and Constrained Control - Applications, Synergies, Challenges and Opportunities.},
      • year = 2024,
      • month = jun,
      • url = {}
      • }
  • MERL Contacts:
  • Research Areas:

    Control, Dynamical Systems, Optimization, Robotics


For automated driving, decision making determines the next maneuver that the vehicle should execute, for which the motion planner will generate a trajectory. The feasibility of the maneuver depends on the current conditions of the vehicle, the route, and the traffic. Thus, decision making must determine which maneuvers are feasible with relatively simple calculations, so that the motion planner, which performs more time-consuming calculations, can succeed in computing the trajectories that achieve the corresponding goals. We propose an approach to solve the decision making problem based on ideas from the reference governor. Our method constructs backward reachable sets for goals and collision areas for maneuvers that are generated by dynamical models parametrized by target values of vehicle motion quantities. Online, the reference governor determines the existence of parameter values that provide membership of the state-parameter vector in a goal reachable set, and non-membership in all collision reachable sets. The resulting online computations are simple and fast, allowing solution of the decision making process at higher rate and with minimal resources as required for standard automotive computing platforms. Furthermore, the method can provide reference maneuvers to guide the motion planning in determining the actual trajectory, can include robustness metrics, and is extended to handle uncertainty in the motion of the obstacles to be avoided. We show simulation results in scenarios involving lane change, braking at intersections, and obstacles with changing velocity.