Recognition and Following of Dynamic Targets by an Omnidirectional Mobile Robot using a Deep Convolutional Neural Network

Abstract

Recent advances in deep learning have stimulated the research related to the application of deep neural networks to solve various problems in robotics. In the field of service robots recognition and tracking of dynamic objects plays an important role since it is crucial for developing behaviors allowing robots to co-exist with humans and other autonomous machines in shared environments. It can simplify the design of autonomous navigation and obstacle avoidance algorithms as well as the ability to operate within multi-agent formations. In this investigation, a deep convolutional neural network is implemented to allow an omnidirectional mobile platform, equipped with onboard TV camera, to recognize and follow other mobile robots moving in the lab. This task can be regarded as a substantial step on the way of achieving our goal to design a dynamic target following behavior for a service robot. During the conducted experiments, the implemented algorithm, based on a deep learning neural network, is able to recognize and localize on a sequence of images the other moving in the lab mobile robot. The omnidirectional mobile platform begins then to follow the detected robot by maintaining a pre-specified relative position and orientation.