Future autonomous agricultural and forestry mobile machines will be capable of operating in an unstructured and partially dynamic environment in a safe and meaningful manner and simultaneously work towards given mission objectives without being extensively controlled by human operators. To do that an autonomous field and service robot must not only localise itself (GIM-Locator) but also sense, perceive, classify and model what is currently happening around it and even predict what will most likely happen in the near future (GIM-Observer). The current wide range of high-quality sensory systems available commercially-off-the-shelf together with well-matured multi-sensor fusion algorithms provide a solid foundation for that task. In addition, the machine should be connected to any available supportive infrastructure (e.g., localization services, dynamic maps, etc.) and have tight interactions with other autonomous mobile machines sharing the same operation space.
Forests are the most important natural resources in Scandinavia. The traditional machine chain in Nordic Cut_To_Length (CTL) forest harvesting system consists of a one-grip-harvester for felling, debranching and cutting the stems to logs, and a forwarder for transporting the logs from forest to roadside in separate heaps for each sorts of timber. Forest trucks transport the timber to sawmills and factories. The productivity of a machine chain is currently greatly defined by the experience of the operators of the forestry machines. In addition, there are several parts of the working process, which are repetitive in nature but still require even the most experienced operator’s attention, when he is trying at the same time to do the most challenging tasks. Therefore, there are substantial economic advantages when the level of autonomy is increased, and assistive operations are successfully implemented.
Current autonomous mobile robotic solutions can already perform several classical agricultural tasks, such as pruning, planting, harvesting, monitoring and others. To do that, robots need to be able to localise themselves and often, also map the environment (GIM-Mapper). In most of the open field agricultural applications, the localisation is naturally based on Global Navigation Satellite System (GNSS), but in some agricultural and in most of the forest environments, satellite signals just cannot be trusted. Those situations are the ones where our feature-based GIM-Locator excels.