M3 Master Deep 3D Pathfinding

Team

  • Daniel Wunderlich
  • Florian Wiese
  • Steven Behm
  • Mareike Glock
  • Paul-Eric Lange

Supervision

André Selmanagic, Dara Khajavi

Core Technology

All machine learning related aspects of our project are built upon Python envionments that were managed with Conda. The reinforcement learning approach uses the TF-Agents library, which is as part of the Tensorflow framework. Our generative adversarial network was created with PyTorch. The machine learning process was accelerated using Nvidia's CUDA toolkit. The training data generation and result visualization was achieved in Unity with C#.

Teamwork

To efficiently work as a remote only team, we used Zoom for our meetings and Slack as our text messenger. We documented our progress and ideas using Confluence and chose Gitlab for our source code version control.

Editors

For pyhton programming we chose to work with Visual Studio Code and it's Jupyter Notebook support. When working on our Unity projects we went with the suggested editor Visual Studio.
The logos were sourced from official style guides and press kits