Reinforcement Learning for Container Terminal

 Overview

Company Description

As a European leading provider of industrial technology solutions, our 1,900 dedicated technical professionals are ready to provide our clients with high-quality services such as recruitment and staffing, training, project-based solutions, consultancy and managed services as well as in-house developed cloud-based products.

 With a track record of over 40 years, ICT Group has both extensive multi-domain expertise and in-depth industry knowledge. We deliver proven technology solutions and are uniquely positioned to help our clients make their business processes more efficient, flexible, simple, secure and – as a result – more sustainable.

Project Description

In this project, you will implement a policy gradient method, for instance Proximal Policy Optimization (PPO), or a variant of policy gradient methods. Policy gradient methods and reinforcement learning methods in general, are notoriously sensitive to preprocessing choices and parameter tuning, making implementation of these algorithms difficult.

Goals of the Project

Container terminals are a vital part of the supply chain. Many optimization problems arise on a container terminal, this project will focus on the stacking problem.

Deliverables

A policy gradient method, for instance Proximal Policy Optimization (PPO), or a variant of policy gradient methods. Policy gradient methods and reinforcement learning methods in general, are notoriously sensitive to preprocessing choices and parameter tuning, making implementation of these algorithms difficult. Next to that, you will perform research on enriching these methods with our extensive domain knowledge, therefore improving the performance of these reinforcement learning algorithms. This project can be implemented in for instance Python, Java or C++.

Essential Student Knowledge

AI, Reinforcement learning, PGM

More information: escf@tue.nl