Volvo Virtual Platform

Overview

Volvo utilizes a simulation program called Vehprop to emulate freight trucks traveling on a route that connects the GBG Harbour to Viared Boras in Sweden. The simulation allows Volvo to virtually test and collect performance data of the trucks prior to actually traveling the road. Vehprop can be used to collect data such as fuel consumption, speed, and temperatures.

Objectives

The vehicle should only use the service brakes only 20% of the time over the entire distance. Reinforcement learning will be implemented to help the program learn and optimize the best solution to minimize the service brake usage and increase the electric brake usage.

Approach

-Describe the approach your team took to solve the problem using a bulleted list of steps
-Understood the problems and solutions Volvo tasked us with
-Learned the current version of Vehprop and how it functions
-Learned Simulink and how it works
-Understood what reinforcement learning is and how it works
-Researched different methods of reinforcement learning
-Researched reward functions in other systems
-Generated ideas where to implement the reinforcement learning module
-Brainstormed the reward functions
-Tested and edited the reward functions
-Researched secure ways of sharing files to external vendors using Python

Outcomes

-The sponsor will be able to run, test, and verify the optimum usage for service brakes and electrical brakes. With Reinforcement learning, the program can teach itself and find the optimum balance between the brake usages.
-The sponsor may continue to research and implement a secure way of sharing files externally.