Motion planning for manipulator in dynamic environment using RGB-D sensor

Abstract

The presented dissertation deals with the topic of trajectory planning of a manipulator in a dynamic environment using data from the RGB-D sensor. An example of such a dynamic environment is a shared workspace where a robot interacts with human workers. Cooperation between robot and human is a widespread topic within the concept of Industry 4.0, which opens up the possibility of creating workplaces with robots that can come into direct contact with employees during the work cycle. Such collaboration brings new opportunities to improve ergonomics and options for manufacturing automation. However, it also carries the risks associated with the possibility of a robot colliding with a human. Adaptive behaviour of the robot - replanning the trajectory with respect to the operator’s current position - can increase the efficiency and safety of cooperation since the robot will be able to avoid collisions and proceed in task completion. In such a situation, however, the user cannot know in advance what the robot’s trajectory will look like after replanning, which can cause discomfort along with reducing efficiency when interacting with the robot. The research in this work focuses on the topic of motion planning and communication of the robot motion plan to a human worker during cooperation in a shared workspace. The requirement is not only a theoretical exploration of possibilities but also a practical implementation in the form of an experimental workplace to verify the proposed principles. The introductory part of the thesis analyses the current state of the art in the field of path planning, motion planning frameworks, environment perception, approaches to improving mutual awareness during human-robot cooperation and implementation of haptic feedback devices. The main contribution of the research is the concept of a novel collaborative system, which combines rapid robot path planning with the system for notifying the user about the currently planned path of the robot and its status. The principles of the system are implemented and tested on an experimental workspace. The robot’s path planning system is based on a motion planning framework optimized for better performance in a set of tasks simulated in a virtual environment. It is hypothesized that the use of the proposed notification system during human-robot collaboration will improve the overall performance, awareness about the planned robot trajectory and encourage a positive experience to the human user. In order to test this hypothesis, a user study is performed, and its data are statistically analysed. The results indicate the potential of the developed haptic notification-based approach in improving mutual understanding during human-robot interaction. The topic of the work is relevant for the deployment of collaborative robots in industrial tasks and aims at improving the effectiveness of human-robot cooperation.