Speaker
Description
Forestry operations are facing increasing structural challenges, including a shrinking skilled workforce, heightened safety requirements, and rising expectations for productivity with reduced environmental impact. These create a clear need for automation technologies that can support or eventually perform demanding tasks autonomously. In 2021, our AORO platform marked a significant milestone as the first fully autonomous unmanned forestry machine demonstrated live. While AORO successfully achieved autonomous navigation and partial crane automation in a forwarding task, it lacked the capability to autonomously complete the full log handling cycle. In particular, the crane could not operate all degrees-of-freedom autonomously due to the absence of grapple sensing and a dedicated vision system, restricting the system’s ability to execute a complete log handling cycle that required complex motions of the grapple.
This work presents a new add-on system that closes this gap and enables full control over all crane’s degrees-of-freedom. The solution extends the AORO machine with an integrated system that combines log detection, precise crane control, and coordinated motion planning. Together, these components allow the machine to perform a complete work cycle: navigate to a work area, identify scattered logs, grasp them securely, and pile them in an organized manner along the roadside.
To perceive the work environment, a dedicated stereo camera is mounted on the crane, providing a clear view of the logs within the operating range. A Convolutional Neural Network (CNN)-based detection model was trained to recognize and localize individual logs under realistic field conditions, including variable lighting and cluttered backgrounds. Once detected, the position of each log is estimated relative to the crane base using calibrated geometric transformations. This ensures that the crane can directly act on the detected targets without additional adjustment.
To achieve reliable and precise handling, the crane control system has been upgraded to operate all joints simultaneously. The grapple has been equipped with new sensors to provide direct feedback: a wired encoder measures rotation, and a wireless sensing solution monitors claw position. These additions significantly improve grasp accuracy and allow the crane to coordinate its movements more effectively during both pickup and placement.
A new motion planning approach based on S-curve velocity profiles coordinates all crane joints throughout each movement. This ensures smooth acceleration and deceleration, reducing abrupt motions that can destabilize logs or impose unnecessary stress on the machine. The result is controlled, efficient motion both when approaching a log and when transporting it to the piling location, contributing to stable handling and reduced mechanical wear.
The integrated system was validated in a public demonstration in which the unmanned machine autonomously navigated along a forest road, detected randomly scattered logs, grasped them, and piled them to one side of the road. The full cycle, from perception to final placement, was executed without human intervention. The demonstration confirmed robustness under realistic field conditions and provided tangible proof of system effectiveness, closing a critical gap in AORO’s manipulation autonomy.
| Keywords | smart-crane-control; computer-vision; machine-automation; forestry-robotics |
|---|
