Modeling of head orientation for applications in manipulator control
(English)Manuscript (preprint) (Other academic)
Many heavy equipment manipulators, such as those found on excavators and forestry machines, currently require continuous manual control. While full automation of these systems is complicated due to the unstructured environments in which they are used, semi-automation can be a useful tool for improving both efficiency and working conditions. In the case of forestry machines, we have located specific work tasks where recent advances in time efficient trajectory planning could be used to increase productivity. However, in order to fully take advantage of these methods, a user interface must be developed that will let the operators communicate their intentions to the system in an intuitive and efficient way. We therefore propose a novel interaction method where headtracking is used to estimate the operator’s gaze in order to specify a target point as input to the trajectory planner for the hydraulic manipulator. In this paper, we investigate the feasibility of such a system, through analysis of task-specific requirements. From dedicated user experiments done in a lab we have identified and tested a model from head orientation to gaze direction, which allows us to predict the gaze angle with an average accuracy of close to 4 degrees. In a field test we have verified that the method is transferable to a real-world setting with comparable results.
Computer Vision and Robotics (Autonomous Systems)
IdentifiersURN: urn:nbn:se:umu:diva-89066OAI: oai:DiVA.org:umu-89066DiVA: diva2:718365