Informational Framework for Minimalistic Visual Odometry on Outdoor Robot

In an unknown environment, assessing the robot trajectory in real time is one of the key issues for a successful robotic mission. In such an environment, the absolute measurements, such as the GPS data, may be unavailable. Moreover, estimating the position using only proprioceptive sensors, such as encoders and inertial measurement units (IMUs), will generate errors that increase over time. This paper presents a multisensor fusion approach between IMU and ground optical flow used to estimate the position of a mobile robot while ensuring high integrity localization. The data fusion is done through the informational form of the Kalman filter, namely, information filter (IF). A fault detection and exclusion (FDE) step is added in order to exclude the erroneous measurements from the fusion procedure by making it fault tolerant and to ensure a high localization performance. The approach is based on the use of the IF for the state estimation and tools from the information theory for the FDE. Our proposed approach evaluates the quality of a measurement based on the amount of information it provides using informational metrics such as the Kullback–Leibler divergence. The approach is validated on data obtained from experiments performed in outdoor environments under various conditions, including high-dynamic-range lighting and different ground textures.