Mobile Robot Localisation for Indoor Environments Based on Ceiling Pattern Recognition In this paper a multi-modal localisation system, that estimates a robot position in indoor environments using only on-board sensors, namely a webcam and a compass, is presented. Ceiling lights are used as beacons. Their position is previously known or self-learned during normal operation. Markov Localisation (ML) is both simulated and experimentally validated. For the prediction step it combines IMU (Inertial Measurement Unit) data and image parameters to compute the attitude of the robot. The update step is then calculated by measuring the distance to possibly visible ceiling lights. The experimental validation of the proposed solution shows that the robot position estimate converges to its real position and the error is kept within decimetres of magnitude.