Event
Patchwork: Domain Randomization Simulation Shared Templates Reduce Memory
1 reports · 1 sources
Zixu Hao and co-authors propose and implement Patchwork, a resource-sharing system for domain randomization simulation. It represents environments as E=T+Δ: shared templates hold common resources, and patches record the changes needed to construct a target environment. The authors implement the system on robosuite and MuJoCo and evaluate it on LIBERO-plus and RoboCasa365 workloads. At 32 replicas, shared-template execution reduces CPU RSS by 88–92% compared with one process per environment. The authors state that controlled experiments show that, within the supported sharing boundaries, memory growth depends more on template diversity and unique asset content than on the number of instances.
Report timeline
- Patchwork: A Resource-Sharing System for Domain-Randomized Embodied Simulation
The authors propose Patchwork, a resource-sharing system for domain-randomized simulation built on the E=T+Δ representation, which uses shared templates plus patches to record environment changes, implemented on robosuite and MuJoCo. On LIBERO-plus and RoboCasa365 workloads, at 32 replicas, shared-template execution reduces CPU RSS by 88-92% compared with process-per-environment execution. Experiments show that memory growth depends more on template diversity and unique asset content than on the number of instances.
arXiv cs.RO ·