| Company | Australian Centre for Robotics (UoS) |
| Country | Australia |
| Website | https://www.sydney.edu.au/engineering/our-research/robotics-and-intelligent-systems/australian-centre-for-robotics.html |
| Category | Planning & Inventory › Remote sensing & mapping |
Forestry: multi-sensor fusion, computer vision, machine learning, mapping, LiDAR, hyperspectral
Robotics research in the forestry sector encompasses fundamental and applied scientific investigation aimed at developing robotic systems capable of reliably performing forestry tasks in complex, unstructured natural environments. This research spans multiple disciplines: mechanical engineering for the design of chassis, manipulators, and end-effectors suited to forestry tasks; computer science and artificial intelligence for perception, planning, and decision-making; electronics for sensor integration; and forest science to define task requirements, quality standards, and operational constraints. Core research problems include reliable autonomous navigation on uneven terrain with soil conditions ranging from firm mineral soils to soft wet peat, accurate detection and classification of trees and terrain features using sensors such as LiDAR, RGB and multispectral cameras, and radar, and robust manipulation of variable natural objects. Research programmes range from early-stage theoretical work on algorithms and hardware concepts through to field validation trials in operational plantation and native forest environments. Major international research hubs include the Skogforsk institute in Sweden, the Finnish Natural Resources Institute (Luke), Oregon State University, and ETH Zurich, each contributing to different aspects of forest robotics from harvesting to planting and monitoring. In Australia, robotics research relevant to forestry has been conducted under the Australian Centre for Field Robotics (ACFR) at the University of Sydney, QUT's Centre for Robotics, and through FWPA-funded industry research projects. Key metrics that forest robotics researchers use to evaluate progress include task cycle time, defect rates, terrain-negotiability limits, system reliability (mean time between failures), and the ratio of human supervision time to productive machine time, all of which must approach or match the performance of conventional systems to drive commercial adoption.
A promising research-driven approach that could enhance productivity and reduce risks through automation and monitoring, but its suitability is limited by a lack of demonstrated, large-scale operational results.
| Field | Value |
|---|---|
| Company | Australian Centre for Robotics (UoS) |
| Country | Australia |
| Type | Company |
| FWPA RD&E | 4.4 |