Reproduce “SAGE: A Split-Architecture Methodology for Efficient End-to-End Autonomous Vehicle Control” 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
Arnav Malawade, Mohanad Odema, Sebastien Lajeunesse-DeGroot, Mohammad Abdullah Al Faruque
SAGE reports 55.66% lower edge energy for three high-definition camera inputs than edge-only execution. It splits an imitation-learning model at an early bottleneck, sending a compact representation to the cloud when conditions favor offloading, or running locally otherwise. The result was evaluated on PX2, not in closed-loop driving, and does not apply to every model; CarlaNet is an exception.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.