Path planning algorithms for terrain navigation are often computationally heavy and rely on categorizing the terrain, which results in limited solution spaces and hinders their real-time implementation.This paper presents a path planning method based on trajectory rollout and model-based forwardsimulation. The path planning algorithm follows a predefined global path with soft constraints forobstacle avoidance. The rollout algorithm proposes multiple candidate paths, which are evaluatedbased on the dynamics and geometries of the machine and the ground. The ground geometry in frontof the machine is dynamically recognized from on-board sensors. The algorithm is evaluated in bothsimulations and physical experiments on a full-scale forest machine. The results show that the pathplanner has a large success rate of finding a feasible path to the goal and is robust against sensornoises and unexpected disturbances in the machine and the forest environment.

Forest machines travel over rough and rocky terrain and the drivers suffer intensive whole body vibrations. Moreover, the heavy weight of the machine often damages soils and plants in the nature environment, which makes the forest industry less sustainable. Autonomous forest machines that can evenly distribute ground pressure and be remotely supervised are emerging solutions. The development of such vehicles requires not only simulations but also physical experiments. A scaled-down test platform emulating the real forest machine significantly reduces the development cost and improves safety. The machine studied in this paper is an articulated forwarder with six wheels mounted on six pendulum-arms. The angles of the pendulum-arms are controlled individually to balance the tire pressure and keep the chassis horizontal. To date, there is no affordable test platform with the same vehicle design. This paper elaborates the mechanical structure and electronics of the design of such a scaled-down test platform. The test platform is reconfigurable in both mechanical structure and the electronics. Its chassis has three sections connected by two actuated joints. Through the control on the two joints, the vehicle can be configured as two or three body sections. The usage of ROS middleware and the design of the electronic architecture allow easy reconfiguration of perception sensors and control actuators. The test platform uses a master-slave architecture of embedded controllers. The master controller is an on-board embedded computer running Linux. The slave controllers implement dedicated functions on perception, control and communication. The platform has a dedicated router to enable remote access to the system. Each pendulum-arm is controlled by a linear actuator and each wheel is controlled by a BLDC motor. The positions of the six arms are simultaneously coordinated by the master controller, so that the vehicle can travel on uneven terrain with higher speed, less cabin vibration, and a more horizontal chassis. The contribution of this paper is the detailed instruction on the design and manufacturing of a 1:5 scaled-down model of a forest forwarder, which can autonomously navigate in a forest environment.

Topography estimation is essential for autonomous off-road navigation. Common methods rely on point cloud data from, e.g., Light Detection and Ranging sensors (LIDARs) and stereo cameras. Stereo cameras produce dense point clouds with larger coverage but lower accuracy. LIDARs, on the other hand, have higher accuracy and longer range but much less coverage. LIDARs are also more expensive. The research question examines whether incorporating LIDARs can significantly improve stereo camera accuracy. Current sensor fusion methods use LIDARs' raw measurements directly; thus, the improvement in estimation accuracy is limited to only LIDAR-scanned locations The main contribution of our new method is to construct a reference ground plane through the interpolation of LIDAR data so that the interpolated maps have similar coverage as the stereo camera's point cloud. The interpolated maps are fused with the stereo camera point cloud via Kalman filters to improve a larger section of the topography map. The method is tested in three environments: controlled indoor, semi-controlled outdoor, and unstructured terrain. Compared to the existing method without LIDAR interpolation, the proposed approach reduces average error by 40% in the controlled environment and 67% in the semi-controlled environment, while maintaining large coverage. The unstructured environment evaluation confirms its corrective impact.

Sustainable forestry requires efficient regeneration methods to ensure that new forests are established quickly. In Sweden, 99% of the planting is manual, but finding labor for this arduous work is difficult. An autonomous scarifying and planting machine with high precision, low environmental impact, and a good work environment would meet the needs of the forest industry. For two years, a collaborative group of researchers, manufacturers, and users (forest companies) has worked together on developing and testing a new concept for autonomous forest regeneration (Autoplant). The concept comprises several subsystems, i.e., regeneration and route planning, autonomous driving (path planning), new technology for forest regeneration with minimal environmental impact, automatic plant management, crane motion planning, detection of planting spots, and follow-up. The subsystems were tested separately and integrated together during a field test at a clearcut. The concept shows great potential, especially from an environmental perspective, with significantly reduced soil disturbances, from approximately 50% (the area proportion of the area disturbed by disc trenching) to less than 3%. The Autoplant project highlights the challenges and opportunities related to future development, e.g., the relation between machine cost and operating speed, sensor robustness in response to vibrations and weather, and precision in detecting the size and type of obstacles during autonomous driving and planting.