EKF-SLAM with Autonomous Exploration using a Low Cost Robot

Autonomous exploration and SLAM (Simultaneous Localization and Mapping) are two of the main problems in robotics due to their relevance for many military and commercial real world applications. This article proposes a solution, tested using a low cost robot (less than US$ 500), based on EKF-SLAM (Extended Kalman Filter) to autonomously explore an static and unknown 2D environment. This environment is assumed to be composed of walls and cylinders which are modeled respectively as line segments and circles. The proposed solution applies border detection techniques to occupancy grid maps and performs path planning using the $\mathrm{A}^{*}$ search algorithm to generate safe and feasible paths for the robot autonomous exploration. A differential drive robot with a Raspberry Pi 3 Model B+ single board computer, a laser scanner, a electronic compass and wheel encoders was designed and built to test the proposed solution. This robot, via a Wi-Fi network, communicates with a remote personal computer that runs the autonomous exploration and EKF-SLAM algorithms. Simulation and real world experiments are shown for an indoor closed environment with approximately 41 square meters, 6 thin cylinders and 10 walls. In the real world experiment the proposed solution achieved an average error of 8.11 cm for the walls’ characteristic points and 4.33 cm for the cylinders’ centroids.