[6-1015-D] Other Categories (3)

[6-1015-D-05] Autonomous Navigation and Obstacle Avoidance for a Robotic Mower using Machine Vision

*Kosuke Inoue¹ (1. The University of Tokyo(Japan))

Keywords:Autonomous Navigation, Visual SLAM, Obstacle Avoidance, Deep Learning, Object Detection, Stereo Camera

The autonomous operation of agricultural machinery using global navigation satellite system (GNSS) information has recently experienced rapid development as a labor-saving measure in agriculture. The self position is recognized with a GNNS signal, and the vehicle can travel in the area autonomously. However, if the vehicle is driven using only the GNSS signal such that the surrounding environment is not recognized, there is a risk of collision with an obstacle. Furthermore, sensors such as radars or lasers cannot distinguish between grass and obstacles and thus cannot be used to detect the likely obstacles encountered by agricultural machinery. Autonomous driving cannot be performed in environments where the satellite positioning accuracy is low, such as orchards. Herein, an autonomous driving system was developed that performs obstacle avoidance and autonomous driving without a GNSS signal by using an object detection system that is based on a stereo camera and deep learning. Stereo cameras and convolutional neural networks recognize the environment and avoid obstacles. The self position is corrected by observing a landmark in the environment. The experiment will be conducted at the Tanashi Forest of the University of Tokyo to evaluate autonomous driving by employing real-time kinematic-GNSS to measure the true values.