Seite - 193 - in Proceedings - OAGM & ARW Joint Workshop 2016 on "Computer Vision and Robotics“

ControllingandTrackinganUnmannedGround VehiclewithAckermanndrive EugenKaltenegger1,BenjaminBinder1,MarkusBader2 InstituteofComputerAidedAutomation ViennaUniversityofTechnology,Austria Abstract This work presents a tracking and control mechanism for an UGV (Unmanned Ground Vehicle) and its integration into ROS (Robot Operating System). The overall goal of which this work is part, is the creation of a fleet of ackermann robots to conduct studies in the field of autonomous driving. In order to achieve this goal a 1:10 RC-race car model is equipped with an Arduino board to control thevehiclesactuatorsandaRaspberryPi tohost theROSserver. Inaddition,aphysics simulation is used to model this car for testing. The shown results support the used velocity motion model and the applicabilityof thedeveloped interface tocontrolboth platforms. 1. Introduction During the lastyears,manystudieshavebeenconducted in thefieldofautonomousdriving[6,1]and the automotive industries as well as companies like Google are showing great interest in this market. Up to 2007, competitions like the DARPA Grand and Urban Challenge pushed research towards autonomous cars with great success [4]. Nowadays, events like the Freescale Cup1, the Carolo-Cup2 and others are created to target young students by using RC-race car models which in terms of costs are very attractive. With this in mind, the Institute of Computer Aided Automation at the Technical University of Vienna is planning to create a fleet of autonomous ackermann robots to attract students and toat onepoint takepart in sucha competition. Thispaperdescribes thecreationof thefirstof thesevehiclesanditssimulationwhilealsointroducing acommon interfaceanda trackingsystemsupporting them. The robot is based on a RC-race car with an ackermann steering. The computation and controlling of the robot is achieved through a Raspberry Pi and an Arduino Uno equipped with a motor-shield. Inaddition thevehicle is simulatedwithGazebo[8], anopensourcesoftware forphysical simulation basedonODE(OpenDynamicsEngine). Tokeepthevehicleandthesimulationcompatible, thesame interface is used, which is based upon the open source software ROS (Robot Operating System) [2]. The vehicle and the simulation are both using the same velocity motion model [4] which equals the prediction step of a Kalman filter for motion tracking [4, 7]. The velocity motion model introduced by Thrun is defined for differential drive robots, but given several changes, which will be further explained, it can also be used for ackermann robots. Since they are commonly used in robotics and provideenoughinformationforanackermannrobotsmotion,differentialdrivecommandsarechosen as input. A ROS node transforms the differential drive commands to ackermann commands, which include a velocity and a steering angle. The robot and its simulation publish their estimated pose and its uncertainty into ROS topics. This allows for easy comparison of the trajectory driven by the 1Freescale Cup: https://community.freescale.com/docs/DOC-1284 (25.04.2016) 2Carolo-Cup: https://wiki.ifr.ing.tu-bs.de/carolocup/ (25.04.2016) 193