Abstract
Autonomous navigation in complex indoor environments remains a significant challenge due to the sparsity of rewards and the high dimensionality of sensor data. In order to enable a Pioneer 3-DX mobile robot to navigate a six-room environment, this work provides a Deep Reinforcement Learning (DRL) technique using the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm. The robot learns to accomplish randomly generated goals while avoiding obstacles like tables and chairs by combining visual input via a Convolutional Neural Network (CNN) with scalar state information via a Multi-Layer Perceptron (MLP). Our findings show how well the suggested reward function and architecture work to achieve reliable navigation in multi-room arrangements, as measured by Success weighted by Path Length (SPL) and Success Rate.