BIFROST: Bridging Invariant Feature Representation for Observation-space Sim2Real Transfer
This work addresses the sim2real transfer problem for robot policy learning by exploiting shared structure between simulation and reality, enabling zero-shot transfer without separate adaptation modules.
BIFROST learns a shared latent representation from paired cross-domain data using a cross-domain bisimulation objective, enabling zero-shot transfer of robot policies from simulation to reality. It achieves effective transfer on sim2sim visual navigation and sim2real manipulation tasks where domain adaptation and co-training baselines fail.
Sim2real transfer for robot policy learning suffers due to mismatch between simulation and reality. Existing methods typically address each gap in isolation through separate adaptation modules, which are composed or layered when both gaps coexist. Yet the basis for attempting sim2real in the first place is that there is shared structure between a task in simulation and reality, where equivalent actions from equivalent configurations produce equivalent long term outcomes regardless of domain specific differences in rendering or physics. In this paper, we study whether we can identify and exploit this shared structure from raw observations to train a policy that enables zero shot transfer. We introduce BIFROST, which learns a shared history encoder on paired cross-domain data via cross-domain bisimulation objective: observation-action sequences leading to equivalent long-term behavior are mapped to nearby latent states, regardless of domain. Policies trained on these latent states in simulation transfer zero-shot to reality. We provide empirical evidence on sim2sim visual navigation and sim2real contact rich manipulation task and visual servoing task that BIFROST achieves effective transfer where domain adaptation and co-training baselines fail under both visual and dynamics domain gaps.