Reproduce “EDEN: Entorhinal Driven Egocentric Navigation Toward Robotic Deployment” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Mikolaj Walczak, Romina Aalishah, Wyatt Mackey, Brittany Story, David L. Boothe Jr., Nicholas Waytowich, Xiaomin Lin, Tinoosh Mohsenin

Developed by Johns Hopkins University and the U.S. Army Research Laboratory, EDEN is a biologically inspired navigation framework for robots that integrates learned grid cell representations with deep reinforcement learning. This framework achieves 99% success in simple navigation scenarios and over 94% in complex environments, demonstrating enhanced robustness and efficiency compared to traditional methods.

Key insights

  • Integrating biologically inspired grid cell representations into deep reinforcement learning can provide structured internal spatial awareness, reducing reliance on external maps.
  • Learning robust spatial encodings through supervised pre-training before reinforcement learning can lead to more efficient and generalizable navigation policies.
  • The use of an internal "cognitive map" based on grid cells enhances resilience to sensory noise and partial observations, addressing limitations of traditional SLAM and pure RL.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.