Derivation of an Optimum and Allowable Range of Pan and Tilt Angles in External Sideway Views for Grasping and Placing Tasks in Unmanned Construction Based on Human Object Recognition

Abstract

Herein, unmanned construction technology that allows remote control of construction machinery has been introduced for post-disaster recovery efforts. One of the most significant problems in unmanned construction is that work efficiency is less than half of that in manned operations owing to the lack of visual information. Thus, deriving optimum and allowable external sideway views can lead to higher efficiencies. In this study, we investigate optimum and allowable pan and tilt angles of external views for grasping and placing tasks as these tasks are required in disaster sites. Humans primarily recognize objects from canonical views with the least amount of occlusions, and with an allowable rotation of 30°. Thus, we hypothesize that external views with pan and tilt angles of 90° could be canonical views for grasping and placing; the allowable range of both pan and tilt angles was ± 30°. We conducted experiments using a 1/20 scale model to derive an optimum and allowable range of pan and tilt angles. The experimental results suggest that the optimum pan angle is 90°, and the allowable range for grasping is ± 30°; the range for placing was from 30° to 135°. In addition, the results indicate that the optimum tilt angle is 60°, and the allowable range is (± 30°).