A Memristor-Based Cerebellar PC-DCN Circuit for Robust Neuromorphic Signal Integration in Robotic Applications

Existing neuromorphic hardware faces severe bottlenecks in balancing biological realism, circuit complexity, and power consumption, particularly in robotic control applications. To address these hardware limitations, this paper proposes an innovative memristor-based neuromorphic circuit inspired by the cerebellar Purkinje cell (PC) and deep cerebellar nucleus (DCN) loop. By leveraging the nonlinear characteristics of memristors, the proposed circuit topology reproduces the dual-mode firing patterns of PCs and the multi-signal integration behavior of the DCN, thereby offering a potential approach to alleviating the conventional trade-off between functional fidelity and hardware overhead. Furthermore, we construct a novel hardware-efficient signal interaction paradigm based on the PC-DCN loop, realizing robust and low-power signal processing through a dynamic excitatory-inhibitory integration mechanism. Circuit-level simulation results demonstrate that the system achieves a signal integration stability exceeding 90% and restricts output fluctuations to within 10% under input noise, exhibiting exceptional interference immunity. Compared to conventional multibus control schemes, the dynamic power consumption is reduced by nearly two orders of magnitude. This circuit-level innovation provides a highly efficient and robust hardware foundation for complex robotic perception and rapid response.