Real-time implementation of the cerebellum neural network

摘要

The cerebellum is an important regulatory center for motor and learning in the human brain and its role has increasingly attracted attentions of researchers. Realizing a cerebellar model can be working on a biological time scale is very important both for exploration of mechanisms and practical application of the functions. In this study, we implement a cerebellum spiking neural network with an efficient method on field-programmable gate array (FPGA), which can generate the spiking activities in real time. Based on this, we propose an adaptive feedback control system with the cerebellum model. The dynamic error of robotic arm is taken as the system input and by using the learning mechanism of the cerebellum, the corresponding correction signal can be exported. The results show that this system can eliminate the error and control the robotic arm.