|Ph.D Student||Kahane Boaz|
|Subject||Robot Integrated Construction with Iterative "Sense and|
Act" Real-Time Processing
|Department||Department of Civil and Environmental Engineering||Supervisors||Professor Yehiel Rosenfeld|
|Mr. Garry-Haim Zalmanson|
Any development of autonomous robots for indoor construction tasks faces the common problem of adapting the robot to its changing environment. This research examines a new approach that allows the robot to be positioned at its workstation with a merely rough accuracy of a few centimeters. Depending on the actual robot-position, the system would plan the appropriate manipulator moves. Moreover, the movements of the robot’s end-tool to its final outset position should be directed in real-time by sensing the very close surroundings of the tool. That is to say, with each manipulator movement for placing an element, the system would assure the correct positioning of the element by performing several iterative “sense and act” cycles. This consistent interaction with the close environment would eliminate many errors that may occur in the farther environment, such as location mistakes, mapping errors, manipulator deflections etc.
The thesis presents and details the principles of the suggested approach, and examines its applicability by actual experiments with the task of wall tiling, which is among the most complicated indoor construction tasks for robots. A unique original end-tool was developed, and was tested in full-scale laboratory conditions, using a CCD camera and 5 laser projectors. The experiment results show that the “sense and act” approach is applicable, and that it could be generalized and modified for implementation in other construction tasks, as well. A thorough analysis of the working process depicts the tasks to be performed by the robot and by its human collaborators, setting up the interaction between them. Economic analysis showed that the suggested system would be profitable in a wide range of cases.