Localization for Indoor Applications with a Cheap Sonar by Particle Filter Estimation

Currently there are many robots that take place indoors with both industrial and domestic applications. Here, an accurate localization is essential to provide them with features such as autonomy and artificial intelligence. However, indoor localization is a problem not fully resolved by the scientific community. One problem of indoor environments is that the walls and ceilings block GPS (global positioning systems) signals that can provide position coordinates. Consequently, it is necessary to use feature extraction and/or measurements performed along the trajectory. Unfortunately, problems such as uncertainty arise when any measurement is made. A solution to solve such problems is a probabilistic framework with state control of the mobile robot besides making a correction step following the dynamic system progress. Hence, this work proposes a cheap sonar based on four ultrasonic sensors, to measure the distance to four landmarks on environment. The readings obtained from the sensors, combined with odometry data correspond to entries for a particle filter. Accordingly, the main contribution of this work is the attenuation of uncertainty through a cheap localization system. The algorithm uses sonar like a second point of view of environment and reduces the uncertainty at each iteration of the particle filter.