From 9e03ad68e0aba7c9ea707a7f61ea76970e7366c3 Mon Sep 17 00:00:00 2001
From: Oier Mees Metrics for Real-to-Sim Evaluation
MMRV aggregates the N^2 rank violations by averaging the worst-case rank violation for each policy.
Visual discrepancies between real-world and simulated environments can comprise a distribution shift that adversely affects a learned policy’s behavior, rendering simulated evaluation unreliable. Our goal is to match the simulator