A Real-Coded Dynamic Genetic Algorithm for Optimizing Driver's Model in Emission Test Cycle

This paper proposes a real-coded Genetic Algorithm for finding optimal parameters' values of a driver's model. Driver's models are one of the important requirements in de-signing controllers for vehicle electrification technology. Latest driver's model that includes driver's response delay, driver's response time and the level of driver's skills has been proposed previously. This model contains several critical parameters that need to be optimally determined in accordance with the vehicle's properties as well as the test cycle data. The objective of this research is to search for the near-optimal values of driver's model parameters that minimize the error between the target speed and the model's speed, as well as maximize the smoothness of the driving operation. The algorithm is evaluated using two driving cycle data, Japan's 10-15 mode cycle and Europe's NEDC. The solutions are compared with other existing heuristic techniques. Experimental results demonstrate a good performance and the consistency of the proposed algorithm.