Multi Time-Scale Engine and Powertrain Control for Autonomous Vehicles Via Lagrange Multipliers

Connected and autonomous vehicles (CAVs) have the ability to use information obtained via vehicle-to-infrastructure (V2I), vehicle-to-vehicle (V2V) communication, and sensors to improve their fuel economy through predictive strategies, including velocity trajectory optimization and optimal traffic light arrival and departure. These powertrain control strategies operate on a slow timescale relative to the engine dynamics; hence, assume that the engine torque production is instantaneous. This assumption results in a torque command profile that may lead to engine dynamics constraint violation, actuator saturation, poor tracking performance, decreased efficiency, poor drivability, and increased emissions. To address this issue, a supplemental controller based on an iterative hierarchical model predictive control (MPC) is proposed in this paper. The constraint satisfaction is achieved through a novel two-way communication of the Lagrange multipliers. The proposed methodology is demonstrated on an autonomous diesel semitruck on two maneuvers. Compared to a traditional centralized approach, the proposed method achieves systematic constraints' satisfaction with negligible effect on fuel economy, less than 1%, and significantly improved computation time, more than ten times.