Trajectory data for a CNN controlled mobile robot

This work explores options for unsupervised training of a convolutional neural network (CNN) for navigation of a mobile robot. A simulated training environment was created in which the robot, through random motion, gathered the required data needed for training. Two approaches to training were investigated, either to selective choose the training data from the random set acquired or to consider modifying the network output to favour improved navigation. The data presented shows the trajectories the robot took using the trained CNNs.