Reinforcement learning technique for parameterization in powertrain controls

As climate change looms large, the automotive industry gears up for an Electric Vehicle (EV) transition, together with the clean energy transition, to pull down our net global greenhouse emissions to zero. It becomes the need of the hour to optimize the use of our resources and meet the requirements of time, effort, cost, accuracy and transient performance brought in by the stringent emission norms and the Real Driving Emissions (RDE) test. In this paper, the authors present Artificial Intelligence (AI) techniques to address real-world challenges for accelerated product development in powertrain control. Genetic algorithm (GA) and Reinforcement Learning (RL) based techniques were effectively utilized to parameterize a time varying electromechanical system. The purpose of a controller is to have the actual value equal to the setpoint across various operating points. The control engineer normally achieves it conventionally with an additive offset and multiple multiplicative factors based on error. The operating region is hence divided into multiple learning zones depending on deviations from the setpoint. GA and RL were used to learn the multiplicative factors and the individual weight maps per zone for a two zone and four zone learning strategy respectively. These AI algorithms proved to be good techniques to model the stochastic nature of processes in powertrain development.