Multilevel-based topology design and shape control of multi-agent systems

Abstract

Extensive control approaches have been developed recently for cooperative controls of multi-agent systems. Some typical control approaches for multi-robot systems include leader-follower, behavior-based, and virtual structure methods. Most existing approaches are not specifically designed for large-scale multi-agent systems. Although some approaches (e.g., potential-based method) have been proposed to solve such a problem, these approaches have difficulty in driving multi-agents to form desired shapes. In addition, potential functions must be carefully selected to achieve global convergence. This thesis aims to develop a new potential function-based shapecontrol algorithm that can guarantee the convergence of large-scale multi-robot systems into a desired shape. Furthermore, the control approach is extended for multiple-cell manipulations, which can be used for many biological analyses, such as single cell analysis, cell fusion, and stem cell differentiation. The study includes the following three aspects: First, a desired region where robots are allowed to stay is formulated. This region serves as a basis for the design of a multilevel-based topology. The grouped robots are required to stay as close as possible with respect to the topology center in a level-bylevel manner. With this multilevel-based topology, various formation shapes can be constructed, and the robots can remain in a desired shape according to task requirements. Second, a novel shape controller is developed based on the proposed multilevel topology. Four forces are generated based on potential functions, which can drive robots to move into the desired region and reach their desired levels while avoiding collisions. Collisions against neighboring robots and fixed obstacles are also considered. A shape regulation control force is further incorporated into the controller to address the local minima issue when the robots get stuck at undesired positions. A direct Lyapunov approach is utilized to analyze the stability of the controlled system. Simulations and experiments are performed using commercially available robots. Third, the proposed multilevel-based topology and control are further extended to the automated transportation of multiple cells with an optical tweezer manipulation system. Cell patterning with various shapes is investigated to locate biological cells at suitable positions. Experiments on locating multiple yeast cells to form different shapes are conducted using robotically controlled optical tweezers. The main contribution of this thesis lies in the proposal of a multilevel-based topology design on which the development of a novel shape controller is based. This shape controller aims to position multiple agents accurately within the desired region to construct numerous desired formation shapes. The approach has been successfully applied to the formation control of robot swarms and multiple biological cells.

Date of Award	2 Oct 2013
Original language	English
Awarding Institution	City University of Hong Kong
Supervisor	Dong SUN (Supervisor)