JSAI2019

Presentation information

General Session

General Session » [GS] J-12 Human interface, education aid

[2C1-J-12] Human interface, education aid: design and creation

Wed. Jun 5, 2019 9:00 AM - 10:00 AM Room C (4F International conference hall)

Chair:Kazuhisa Seta Reviewer:Jun Ichikawa

9:20 AM - 9:40 AM

[2C1-J-12-02] Decision Making for Model Based Design by Reinforcement Learning

〇Tatsuhide Sakai1, Takahiro Inabe2 (1. Great Wall Motor, 2. Sakiyomi AI labo)

Keywords:reinforcement learning, model based design, passenger vehicle, deep learning

The Model Based Design is identified that hierarchy is structured on functions of each part to achieve a competitiveness in a product design. As the hierarchy becomes complicated, design variables have a huge data space, so it is difficult to properly make decisions in a short time even if a designer has extensive experience. It is verified whether reinforcement learning is effective for the design of electric vehicles. When applied to the vehicle performance of the top of hierarchy, the design limit of energy consumption was derived from the variables space of 128 to the 17th power and the optimal solution for Package was learned from the variables space of 10 to the 77th power.