Real-time target point identification and automated log grasping by a forwarder, using a single stereo camera for both object detection and boom-tip control

Abstract

The forest industry is constantly striving to increase productivity and cut costs, and many research and innovation projects are currently focusing on semi-automated or autonomous systems. A key element, with several possible solutions, is automated log grasping, where researchers and manufacturers are looking for efficient and sturdy ways to solve the task in real-time forwarding operations. This study presents a simple method for automated log grasping using only a single stereo camera for object detection (log and grapple) and a simple controller moving the boom, with feedback from the camera as boom-tip control. The accuracy, precision, and repeatability of the method was tested on a full-scale forwarder. Boom movements were examined from two different start positions in relation to the target position, with the log placed at three different angles. The overall log-grasping success was also evaluated. The tests were performed in a full-scale, real-time operation, without hand-eye calibration or other sensor data from the machine. The method was precise, with high repeatability, but the grasping point showed a minor systematic offset, depending on log angle. However, the deviation in accuracy was too small to affect the success rate. In practice, the most difficult log angles can be avoided by moving the machine slightly. The log grasping method may become part of an autonomous forwarding system or could provide operator support in semi-automated systems.