Design of a reinforcement learning agent for the operation of a depot for buses with alternative drives

Context

Public transport plays a decisive role in the search for sustainable mobility solutions. Especially in the case of buses, diesel vehicles are currently still in use for the most part and are to be replaced with suitable alternatives within the next two decades. For local public transport operators, this means extensive changes to their depots and operating procedures. Since most of the buses arrive late in the evening and have to be ready to run early in the morning, the charging and refueling infrastructure becomes a bottleneck. The goal of the reinforcement learning agent to be trained is to use the stochastically arriving buses, their states (loading, refueling) and route planning for the next morning, and the available resources (refueling stations, loading bays) to make decisions on how to proceed at the depot and thus enable efficient operations.

Goals of the work

• Creation of the stocahstical profile for buses
• Selection of a suitable cost function
• Training of a RL agent

Requirements

• Experience in Python
• Knowledge of machine learning (reinforcement learning is a plus)
• Motivation and interest in solving technical problems independently