A Mobile Robot with an Autonomous and Custom-Designed Control System

Teaching autonomous mobile robots (AMRs) to acquire knowledge independently has been a formidable challenge, characterized by protracted convergence times and computational intensity within traditional methods. This research introduces an innovative paradigm employing a customized spiking neural network (SNN) to address these challenges, fostering autonomous learning and control of AMRs within unfamiliar environments. The proposed model amalgamates spike-timing-dependent plasticity (STDP) with dopamine modulation to augment the learning process. Incorporating the biologically inspired Izhikevich neuron model imparts adaptability and computational efficiency to the control systems, particularly in response to dynamic environmental alterations. Evaluation through simulations elucidates initial challenges during the training phase, where the infusion of brain-inspired learning, dopamine modulation, and the Izhikevich neuron model introduces intricacies, notably manifesting in difficulties adapting to diverse obstacle scenarios. Initial performance metrics reveal a 73% accuracy rate in reaching the target with a 27% collision rate in single obstacle scenarios. However, progressing to the testing phase demonstrates substantial enhancement, culminating in a remarkable 98% accuracy in reaching the target and a marked reduction in collisions to 2% in single obstacle scenarios. These outcomes underscore the model’s adaptive prowess and proficiency in navigating complex environments with varied obstacles. The innovative application of the customized SNN, integrating STDP and dopamine modulation, showcases promising potential in surmounting the challenges associated with reinforcement learning in AMRs.