New Zealand’s forests are commonly planted on steep slopes and include large trees. The research in this thesis contributes to the long term objective of developing a teleoperated mechanised harvester capable of working on slopes of up to 45° steepness and with large trees. The final aim of the project is to create a forestry harvester that provides a comparable improvement to productivity of steep slope harvesting that has been experienced with mechanisation of flat land forestry harvesting in New Zealand. This thesis studies the feasibility and practicability of teleoperation and autonomous control of forestry harvesters, particularly the predominantly excavator-based forestry harvesting machinery in use in New Zealand. The studies were performed on a laboratory scale hydraulic system, but were not able to be tested in the field, but it is considered that the studies presented have made a significant contribution to retrofitting existing forestry harvesters as used in New Zealand. The forestry industry makes a significant contribution to New Zealand’s economy. A large amount of steep land was converted to plantation forestry in the 1980s, and these forests are coming to harvesting age. Unfortunately, the forestry industry has high rates of deaths and serious injuries due to the hazardous nature of the work. While mechanised harvesting is used where possible to the improve safety and productivity, much forestry harvesting is carried out on steeply sloped land. Steep slope machinery is available in other countries but the large tree size in New Zealand precludes its use here. New Zealand-specific machines for steep slope harvesting are under development, but most steep slope harvesting is still performed by hand. The specific areas of research this thesis addresses are in teleoperation – autonomous control, visual feedback, haptic feedback, as well as the use of minimal modelling for system identification of the hydraulic machinery. Trinder Engineering have developed a steep slope harvester using a system of cable anchoring for safety. While this steep slope harvesting machine can perform its task well, the research on teleoperation is aimed at eliminating the hazard to the worker by removing the worker from the machine on the steep slope – “no worker on the slope, no hand on the chainsaw” – without sacrificing the worker’s productivity. Teleoperation of forestry machinery is a difficult problem, primarily due to the unstructured and uncontrolled environment in which forestry harvesting takes place. However, improving technology is easing the difficulties of implementation and is making teleoperation of forestry machines feasible with off-the-shelf computing and networking hardware. The research in this thesis is aimed at maximising the use performance of off-the-shelf computing hardware and software. The system identification uses minimal modelling to develop the dynamic equations of force and motion. Unlike data driven modelling, which fits the system to the data, minimal modelling is driven by a simplified dynamic model, with inaccuracies being modelled as disturbances. The advantage of minimal modelling is that it is physically derived, so it is easy to see how changes in the machine properties affects the behaviour. This has the potential to allow various types of machinery to be modelled in this way, in particular the minimal modelling was first used to model the dynamic behaviour of a Phantom Omni haptic feedback arm, before being tested on a laboratory scale hydraulic system. This research was used to demonstrate the transferability of minimal modelling between different types of mechanical systems. The emphasis on retrofitting existing machinery in the forestry industry flows through to a strong emphasis on using control systems suitable for retrofit during the research, including for system identification. The hydraulic test system has been used to test autonomous control, including autonomous obstacle avoidance. The objective is to make teleoperation easier, by removing the requirement of operators to operate all the hydraulic rams themselves. The state of the machine was sensed by measurement of ram lengths, which in combination with a model of the machine joints enabled accurate calculation of joint angles. In particular, the key novel contribution to the state of the art is in implementation of semi-autonomous teleoperation control using ROS (Robot Operating System), off-theshelf components and ‘Minimal Modelling’ based system identification. The use of these techniques show it is possible to retrofit autonomous controllers in a modular way, without having detailed design information about the machines or having to design heavily customised control systems. The research finds that autonomous control using ROS and minimal modelling based system identification techniques are useful for retrofitting excavator based forestry harvesters. In particular, the modularity of the ROS based system and the ability of the minimal modelling to be extended to non-linear systems using linear time variant models of damping to account for the effects of friction and actuator dead zone. The autonomous control tests on the laboratory hydraulic system were successful, showing it is possible to control a hydraulic machine using a ‘conventional’ electrically operated hydraulic retrofit to perform autonomous movements. This, along with the strong industrial nature of this research work, is a significant research step forward to development of safer harvesting machines in New Zealand.