Study of Autonomous Robotic Lawn Mower Using Multi-Sensor Fusion Based Simultaneous Localization and Mapping

Most robotic lawnmowers on the market adopt buried metal wires to define the boundary and random walk movement to mow the lawn, leading to higher installation costs, overlapping work paths, and inefficient coverage. This study uses the data fusion of multiple sensors, including actuator encoders, Inertial Measurement Unit (IMU), and Light Detection and Ranging (LiDAR), for robot localization and path planning to develop an intelligent robotic lawn mower that automatically detects the mowing boundary and completely covers the lawn. Extended Kalman Filter (EKF) and Adaptive Monte Carlo Localization (AMCL) are used for robot localization. Gmapping SLAM algorithm is used to build the layered costmap, which serves as the basis for path planning. In response to the possibility of temporary changes to the mowing range, a virtual wall function was introduced to customize the work area. The proposed path planning combines partitioned boustrophedon path planning and boundary following to increase the coverage efficiency. A prototype is constructed, and the experimental result is presented to verify the strategy's feasibility and demonstrate its efficiency. The robotic lawnmower achieves a coverage rate of 92% in the test area of $\boldsymbol{24 \mathrm{m}^{2}}$ in about 20 minutes.