Inspiration
As space vehicles become more advanced, many rely on machine learning solutions to manage the complex operation of the many controls. We took inspiration from the utility of how smart homes use machine learning to adapt to many different speech patterns. More specifically, we were interested in taking advantage of the robustness of AI models
What it does
Our project simulates different space environments to train AI to complete specific tasks
How we built it
We used the Unity Engine as a 3D basis for our physics scenarios and used Unity's Machine Learning Agents Toolkit to simulate space vehicles using deep reinforcement.
Challenges we ran into
We ran into several challenges as first-time, first-year hackers who have never worked with subjects such as AI or Machine Learning: compatibility issues between Python, PyTorch, and ML-Agents; difficulty grasping deep reinforcement concepts; technical difficulties regarding implementing unique physics scenarios with machine learning; training the models to fit our scenarios.
Accomplishments that we're proud of
Ultimately, we're proud of finishing our project and submitting it. We were also successful in implementing our ML into our unique physics scenario.
What we learned
We learned a ton of new skills over the weekend: C#, Unity, Machine Learning, AI, Deep Reinforcement, etc.
What's next for SVS Lunar Client?
Given more time for polish, we plan on fully expanding our models to their fullest potential.
Log in or sign up for Devpost to join the conversation.