In New Zealand, the forest industry is economically significant, representing the third largest export by value. Harvesting and extraction of trees pose serious health and safety issues. These issues have become more pronounced with the prevalence of forestry blocks established on steep slopes. The steep slopes (> 20°) preclude the operation of traditional machines such as harvesters and feller-bunchers. Without mechanization, felling is performed manually by ground crew with hand-held chainsaws. The forest industry now leads fatality and injury statistics with 42% and 35% respectively caused during the felling process. The purpose of this research, therefore, is to design and build a semi-autonomous tree felling robot suitable for the unique forestry conditions found in New Zealand. This is motivated by the desire to reduce the aforementioned fatality and injury statistics. Currently available machinery for forest harvesting is not suitable for New Zealand’s steep terrain. Tracked and wheeled vehicles cause significant damage to the hill side and struggle to provide sufficient traction on slopes. Vehicles with a high center of mass can roll, putting the driver in danger; a problem exacerbated by the slope inclination. Although some machines have been modified for increased stability on steep terrain, their lack of agility and continued damage to the terrain is prohibitive. Therefore, from literature of the state of the art, it is seen that there is no suitable forest harvesting machine compatible with New Zealand’s terrain and the requirements of forest managers. The proposed solution is a system using arboreal locomotion similar to that used by monkeys, resulting in a machine supported in the trees rather than on the ground. This method of traversal between trees has not been used for forestry previously and introduces a paradigm shift for forest planning. Due to the unique motion of this machine, it can traverse over terrain typically impractical for traditional ground based felling systems. In order to provide semi-autonomy and remote operation, a suite of sensors and control systems are required for this new application. A novel tree detection algorithm using laser scanning is proposed as a solution to gaining situational awareness for the environment. This enables the robot to identify neighboring trees and analyze which are suitable for traversal and are ready to be harvested. Identification of objects such as trees, ground and other clutter are used in motion planning in order to avoid collisions. Choosing a suitable route through the forest can be a difficult problem for remote operators. Therefore a route planning algorithm is developed to find suitable routes. The planning algorithm considers the known kinematics of the machine and combines a-priori map information with gathered data to calculate routes through the entire forest to minimize a cost function. The cost function is determined by the forest manager with considerations towards fuel consumption, time and yield for example. This system allows forest managers to maximise the robot’s utility. Finally, to provide tree felling capability, an autonomous felling mechanism is presented. In a novel development, force sensors are integrated into the chainsaw bar. These senors provide real-time feedback to the felling algorithm regarding the downwards force of the tree on the bar. By measuring the force, the state of the tree fall can be determined allowing correct timing to be achieved. This document outlines the advancements to the state of the art through theoretic basis, conceptual development, methodology and results gathered in the laboratory and field. This research work is believed to represent a very significant step forward to achieving a commercially viable, steep terrain tree harvesting robot.