Control system by kinect vision for autonomous mobile robots in autonomous driving task

Abstract

This work describes a developed system for autonomous mobile robots in autonomous driving task. This system contains two sub-systems which are the vision and control system. It was designed aiming to participate on the Autonomous Driving league of the Portuguese Robotics Open. The vision system uses a Microsoft Kinect camera and the obtained information from its double integrated sensor (vision and distance) are fused in order to build a filter. This new approach reduces the computational cost due the filter which keeps only the relevant information. This system is able to recognise the specific traffic signs, obstacle detection and it also provides input information for the control system. The control system receives information from the vision system and the main task of the control system is the steering control required in autonomous driving task. The control system has different approaches, from simple to conventional methods as the Proportional-Integral-Derivative (PID) controller. In order to apply the developed system to the existing robots (Formula UM TD1 and Formula UM TD2) with different hardware architecture, a flexible system was developed. Results were obtained from laboratory experiments and also from the competition where the two robots achieved the second and third places.