clamesc/Training-Neural-Networks-for-Event-Based-End-to-End-Robot-Control

The project is a Master's thesis from circa 2016. While it represents an early and interesting combination of neuromorphic sensing (event cameras) and end-to-end control, it lacks the characteristics of a defensible software product. With 59 stars and 23 forks accumulated over 8 years and a velocity of zero, it is essentially a dead repository serving as a historical reference. The tech stack likely relies on deprecated versions of ROS and early RL libraries, making it difficult to deploy today without a full rewrite. In the current landscape, more robust frameworks for event-based vision (like those from the ETH Zurich Robotics and Perception Group or Prophesee's Metavision) and modern SNN libraries (like SpikingJelly or Intel's Lava) have rendered this implementation obsolete. Frontier labs are unlikely to compete here directly as the niche is highly hardware-dependent and academic, but the project is displaced by newer research and commercial SDKs in the neuromorphic space.