Indoor real-time localisation for multiple autonomous vehicles fusing vision, odometry and IMU data Alessandro Faralli, Niko Giovannini, Simone Nardi, and Lucia Pallottino. Research Center E.Piaggio, Faculty of Engineering, University of Pisa, Largo Lucio Lazzarino 1, 56122 Pisa, Italy {ale.fara.90@gmail.com,nikogiovannini3@gmail.com,nardi@mail.dm.unipi.it, lucia.pallottino@unipi.it} Abstract. Due to the increasing usage of service and industrial au- tonomous vehicles, a precise localisation is an essential component re- quired in many applications, e.g. indoor robot navigation. In open out- door environments, differential GPS systems can provide precise posi- tioning information. However, there are many applications in which GPS cannot be used, such as indoor environments. In this work, we aim to increase robot autonomy providing a localisation system based on pas- sive markers, that fuses three kinds of data through extended Kalman filters. With the use of low cost devices, the optical data are combined with other robots’ sensor signals, i.e. odometry and inertial measurement units (IMU) data, in order to obtain accurate localisation at higher track- ing frequencies. The entire system has been developed fully integrated with the Robotic Operating System (ROS) and has been validated with real robots. Keywords: Localisation indoor, odometry, IMU, EKF, passive marker. 1 Introduction A fundamental problem for an autonomous mobile robot is knowing its current position and orientation by sensorial observation and previous accurate localiza- tion. This is still the subject of several researches in the mobile robot community with the aim of increasing robot autonomy. Although global positioning system (GPS) is suitable for mobile robot localization in outdoor environment, it is diffi- cult to be used in an indoor environment. In case GPS is unavailable, localization using odometry [1] and dead reckoning using IMU sensors [2] may provide an alternative solution. However, odometry is subject to growing errors over time and it is hence insufficient for many tasks [3]. The indoor navigation is based on the exploitation of the environment and of available technologies that allow localisation even in indoor scenarios. One of the most widely used techniques is to place landmarks in known environment’s points. In this way mounting an on- board robot webcam (focused on the landmarks), the localisation system uses