Perception Systems for Autonomous Forest Machinery

A prerequisite for increasing the autonomy of forest machinery is to provide robots with digital situational awareness, including a representation of the surrounding environment and the robot's own state in it. Therefore, this article-based dissertation proposes perception systems for autonomous or semi-autonomous forest machinery as a summary of seven publications. The work consists of several perception methods using machine vision, lidar, inertial sensors, and positioning sensors. The sensors are used together by means of probabilistic sensor fusion. Semi-autonomy is interpreted as a useful intermediary step, situated between current mechanized solutions and full autonomy, to assist the operator. In this work, the perception of the robot's self is achieved through estimation of its orientation and position in the world, the posture of its crane, and the pose of the attached tool. The view around the forest machine is produced with a rotating lidar, which provides approximately equal-density 3D measurements in all directions. Furthermore, a machine vision camera is used for detecting young trees among other vegetation, and sensor fusion of an actuated lidar and machine vision camera is utilized for detection and classification of tree species. In addition, in an operator-controlled semi-autonomous system, the operator requires a functional view of the data around the robot. To achieve this, the thesis proposes the use of an augmented reality interface, which requires measuring the pose of the operator's head-mounted display in the forest machine cabin. Here, this work adopts a sensor fusion solution for a head-mounted camera and inertial sensors. In order to increase the level of automation and productivity of forest machines, the work focuses on scientifically novel solutions that are also adaptable for industrial use in forest machinery. Therefore, all the proposed perception methods seek to address a real existing problem within current forest machinery. All the proposed solutions are implemented in a prototype forest machine and field tested in a forest. The proposed methods include posture measurement of a forestry crane, positioning of a freely hanging forestry crane attachment, attitude estimation of an all-terrain vehicle, positioning a head mounted camera in a forest machine cabin, detection of young trees for point cleaning, classification of tree species, and measurement of surrounding tree stems and the ground surface underneath.

Department dc:contributor.department

Sähkötekniikan ja automaation laitos

Grantor dc:publisher

Aalto University

Year dc:date.issued

2023

Author dc:creator

• Hyyti, Heikki

Advisors dc:contributor.advisor

• Lehtola, Ville, Prof., University of Twente, Netherlands
• Visala, Arto, Prof., Aalto University, Department of Electrical Engineering and Automation, Finland

Contributors dc:contributor

• Aalto-yliopisto
• Aalto University

Language dc:language.iso

en

Repository record dc:identifier.uri

https://aaltodoc.aalto.fi/handle/123456789/123885

OAI record oai:identifier

https://aaltodoc.aalto.fi/server/oai/request?verb=GetRecord&metadataPrefix=dim&identifier=oai%3Aaaltodoc.aalto.fi%3A123456789%2F123885