Fully Automated Implementation of Reservoir Computing Models on FPGAs for Nanosecond Inference Times
Fabian C. Legl, Jonas Kantic · 2024
We propose an efficient and generic Field-Programmable Gate Array (FPGA) implementation of Reservoir Computing using Cellular Automata models for the application of time series processing. Since our implementation only uses lookup tables and registers, it can not only be run on high-performance but also on low-cost FPGAs without special hardware components. Our implementation results from a fully automated model definition to FPGA bitstream design automation process. Hence, it significantly reduces the design time and complexity. The generated implementation is three to five orders of magnitude faster than other Reservoir Computing models. This enables intelligent real-time sensor signal processing for applications requiring MHz inference rates, like structure-borne noise monitoring or high-frequency oscillation analysis.