Neuroevolution Based Optimization of Hybrid Transmission Shift Points

Abstract

In a hybrid vehicle the speeds at which the transmission shifts between gears has a large impact on energy consumption. Dynamic Programming (DP) can be used to find the optimal transmission gear for a given torque and speed. Neuroevolution of Augmenting Topologies (NEAT) is then utilized to find the shift lines which allow the transmission to be at the most energy efficient point in an implementable way. Through this strategy, an improvement of 7% was achieved compared to the traditional approach.