Autonomous Photometric 3D Surface Construction by Potential Field Motion Planning

Dugan Um • Darion Grant • Jeongsik Shin • W.-H. Lee

10:00 - 10:30 | Monday 25 September 2017 | Ballroom Foyer



Photometric 3D surface construction has potential for various applications in many fields. UAVs are dominant in 3D photometry due to ease of path planning and control. An autonomous UGV capable of producing high quality 3D photometry is, however, rare and still a daunting task to build due to complexity in ground structures. The main challenge is to ensure the acquisition of a complete set of photography for a given site, minimizing duplication for less data processing and yet to have enough duplication between photos for the stitching process later. From the path planning perspective, it calls for search completeness of a given connected space and yet to ensure minimal duplication in the path for minimum stitching process time. In addition, continuous photographing motion itself has to be slow enough for the highest quality and yet has to be fast enough to minimize the time of photography. Due to many conflicts in photographing and path planning, autonomous photometric 3D surface construction is a daunting task. Three main technical challenges have been identified toward autonomous UGV based 3D photometry; motion control for high quality photography; path planning for complete 3D surface construction; and localization for avoiding same site visit to minimize data processing. In order to resolve these three issues, we built a robotic system with the capability of stable photography, path control by potential field, and localization by SLAM technique. In addition, we studied the constructed 3D surface quality with respect to several motion control and photography parameters such as overall motion speed, acceleration, exposure timing, and frame interval. The robot is navigating in a given environment via potential field path planning, producing promising results of a full 3D environment reconstruction in a single path. Our approach is simple in that the robot is an agent navigating in a potential field where detected environmental significance provide sources of attractive force, while previously occupied locations estimated by SLAM technique provide sources of repelling force. The collision due to sonar sensor failure is compromised by gyroscope thus the robot moves continuously until completing the 3D mapping process. The proposed mapping technique is useful and easy to implement in various applications including harsh or dangerous environment such as tunneling, nuclear site mapping, structural soundness testing or real estate application, etc.