| Summary: | Mobile robots can be used in domestic, industrial or humanitarian fields. Typically, low-cost mobile robot platforms are equipped with sparse and noisy sensors on board, such as array of infrared sensors. In robotics, the ability to map the surrounding area and determine self-location is essential for autonomous navigation. In this paper, the objective is to develop such capability known as Simultaneous Localization and Mapping (SLAM) algorithm for mobile robots with array of infrared sensors. To improve the robot’s observations from noisy sensor measurements, neural network was used to interpret adjacent sensor measurements into grid cells occupancy. In this grid-based SLAM algorithm, Rao-blackwellized particle filter (RBPF) was integrated with neural network. Two different proposal distributions for RBPF; Gaussian approximation, and two-step sampling, were experimented with and without neural network integration in this paper. The results show that the two-step sampling method with neural network integration gives the lowest error of robot state estimate and the highest score of overall map estimate. This integration of grid-based SLAM algorithm, reduced the pose error by approximately 35% and increased accuracy of overall map estimate by 22%. From the experiments, it is concluded that the grid-based SLAM algorithm integrated with neural network and two-step sampling method is feasible for low-cost mobile robot with sparse and noisy sensor measurements. © 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
|
|---|