🤖 AI Summary
This work addresses the lack of a unified, reproducible framework for developing and evaluating physical reservoir computing (PRC). To this end, we propose OpenPRC, an open-source, end-to-end Python framework that, for the first time, seamlessly integrates GPU-accelerated simulations (e.g., via the demlat engine) with real-world video experimental data through a standardized HDF5-based data architecture. OpenPRC supports comprehensive analysis from physical dynamics to task performance by incorporating physics-aware optimizers, information-theoretic diagnostic tools, and modular learning layers. The framework is compatible with external simulation engines such as PyBullet and has been validated on tasks including origami-based systems. It provides standardized benchmarks, capacity analyses, and correlation diagnostics, thereby promoting rigor, standardization, and reproducibility in PRC research.
📝 Abstract
Physical Reservoir Computing (PRC) leverages the intrinsic nonlinear dynamics of physical substrates, mechanical, optical, spintronic, and beyond, as fixed computational reservoirs, offering a compelling paradigm for energy-efficient and embodied machine learning. However, the practical workflow for developing and evaluating PRC systems remains fragmented: existing tools typically address only isolated parts of the pipeline, such as substrate-specific simulation, digital reservoir benchmarking, or readout training. What is missing is a unified framework that can represent both high-fidelity simulated trajectories and real experimental measurements through the same data interface, enabling reproducible evaluation, analysis, and physics-aware optimization across substrates and data sources. We present OpenPRC, an open-source Python framework that fills this gap through a schema-driven physics-to-task pipeline built around five modules: a GPU-accelerated hybrid RK4-PBD physics engine (demlat), a video-based experimental ingestion layer (openprc.vision), a modular learning layer (reservoir), information-theoretic analysis and benchmarking tools (analysis), and physics-aware optimization (optimize). A universal HDF5 schema enforces reproducibility and interoperability, allowing GPU-simulated and experimentally acquired trajectories to enter the same downstream workflow without modification. Demonstrated capabilities include simulations of Origami tessellations, video-based trajectory extraction from a physical reservoir, and a common interface for standardized PRC benchmarking, correlation diagnostics, and capacity analysis. The longer-term vision is to serve as a standardizing layer for the PRC community, compatible with external physics engines including PyBullet, PyElastica, and MERLIN.