Navigation System for Modular Agricultural Machines using Optimal Control Methods and Industrial Standard Network

In agriculture, there is a growing demand to increase the efficiency of machinery. This has led to increasingly larger machines and an increasing usage of automation. The challenge is that there are numerous different combinations of tractors and implements and they all have to work together. The ISO 11783 standard is a widely accepted standard for realising an information exchange between the tractors and implements. The motivation for this thesis was to reduce the workload of the driver and also to improve the efficiency and accuracy of field work. This was achieved by removing the human being from behind steering wheel and letting the navigation system do the driving. Because of their size, implements are more commonly towed these days. The navigation system should also take into account the towed implement. In most commercial products, only the tractor is considered. Two specific goals were set: 1) to study a combined navigation system that is able to drive at a speed of at least 12 km/h with less than 10 cm lateral error under real field conditions; 2) to discuss and present the ways in which a decentralised and generic combined navigation system can be realised using the ISO 11783 network. The solution was to use Nonlinear Model Predictive Control (NMPC) in path tracking. With NMPC, a controller can be realised that controls the motion of both the tractor and the trailer in an optimal manner. In NMPC, the reference trajectory is usually coupled with time. However, in path tracking for agricultural navigation, it is more suitable not to couple the positions of the tractor and the implement with time. Hence, the NMPC had to be modified. Also, other difficulties that arise from the usage of NMPC had to be overcome: the NMPC needs an accurate estimate of the state, so an optimal state estimator that merges the local and global position measurement was developed. Semi-automatic tuning methods were developed for tuning and configuring the numerous parameters needed for combined navigation. The reference path should be feasible, so a feasible path planning algorithm was developed. Additionally, a collision avoidance method was developed and incorporated within the NMPC. The results of the field tests show that the goals were met. The accuracy of the navigation system is within the given limits. The system is also able to complete the field work without human intervention, including making headland turnings and avoiding electricity poles. The required information on combined navigation has been illustrated based on the case studies. It is possible to build system based on the ISO 11783 standard and distribute sensors and actuators.