Memcapacitor-Based Operant Conditioning Neural Network With Deprivation and Its Application in Inspection Robots

Nowadays, memcapacitor-based associative memory neural networks are focusing on classical conditioning roles and ignoring operant conditioning roles. In this article, a biomimetic model of operant conditioning neural network based on memcapacitor is designed. The designed circuit includes neuron module, time delay module, hunger output module, experience module, and decision making based on experience module. The novel neural network based on memcapacitors implements learning, forgetting, immediate and delayed reinforcement learning, blocking, generalization, and decision making. In addition, the effects of hunger and satiety on operant conditioning are discussed and implemented using memcapacitors to represent states of deprivation. PSPICE simulation results show that the circuit can be used to simulate real-world conditioned reflexes and complex applications. The proposed circuit can be applied to an intelligent inspection robot for power distribution rooms, enabling autonomous learning and equipment detection.