Spiking network; Izhikevich model; STDP; agent

The aim of this paper is a proposal of a spiking neural network circuit design continuously modifying synaptic strengths between neurons through the spike-timing-dependent plasticity to fulfil the agent objective. The network consists of Izhikevich neurons controlling an agent which uses three sensors to associate obstacles to the defined classes in its environment. The environment is formed as a torus of revolution. The reinforcement of synapses of the control network is based on the spike-timing-dependent plasticity (STDP), which is modulated by extracellular dopamine. The controlled agent may perform four different output actions: move forward,turn left, turn right, and release repulse beacon to avoid all obstacles with a negative impact on the agent. The testing environment has been populated with obstacles randomly placed in the torus. Experiments confirmed that the agent is able to associate negative and positive object classes in its environment. The continuation of this work was outlined in the conclusion.