Collision Avoidance Method with Nonlinear Model Predictive Trajectory Control

The automatic navigation systems are able to control the tractor and even the implement without human interaction. However, if there is no device to recognize obstacles on the field, a human driver is still needed to ensure that the tractor does not collide with anything, like electricity poles.In this research, the collision avoidance method was built on top of the existing experimental navigation system which is able to control both the tractor and the towed implement with the help of true MIMO controller. The existing navigation algorithm is based on the Nonlinear Model Predictive Control (NMPC) which is modified to support path tracking. The augmentation of the collision avoidance to the NMPC was inspired by the potential field method. The proposed solution does not increase the computational cost of the original NMPC.The collision avoidance method was tested and was proven to work in real environment at driving speeds less than 3.5 m/s. The obstacles were detected with a 2D laser scanner which was mounted in the front of the tractor. The obstacle detection was also found to be sufficiently accurate to current application.