Laser scanner based collision prevention system for autonomous agricultural tractor
Aalto University, Department of Electrical Engineering and Automation, Otaniementie 17, 02150 Espoo, Finland; e-mail: timo.oksanen@aalto.fi
Abstract:
In manned agricultural vehicles, the automated systems assist the driver by reducing the workload. This is achieved e.g. by using an automatic guidance system to steer the tractor along the desired path. However, increasing automation tends to cause a reduction of awareness, so risks to collide obstacles in the field are higher. In this study, an autonomous tractor was equipped with front side laserscanner (LIDAR) to sense the environment in front. The laserscanner scans the environment at 50 Hz rate. The theoretical maximum range of the sensor is 25 m, but it was found in the tests, that in agricultural field conditions, the feasible range is not more than 7 m, due to the sunlight disturbance. Agricultural vehicles weigh tons, so the deceleration is limited and the limited range causes challenges to detect the obstacle and decelerate without colliding it. The developed algorithm is able to detect solid objects, like electricity poles in the trajectory. The deceleration algorithm is based on the known dynamics and actuator delays of the tractor locomotion system, by taking into account the maximum deceleration rate. In field tests, the system was evaluated in grass fields. In the first test, the system was tested with real electricity poles with no implement. In the second test, the system was tested with a mower and by using artificial obstacles placed into the grass. The system was able to detect the obstacles with high accuracy and stop precisely, but in the corners of the field the system caused false positives when the sensor was sensing beyond the edges of the field plot.
Key words:
accidents, autonomous emergency brake, control systems, environment sensing, robotics, tractors, vehicles