UC Irvine

In this thesis, we introduce an easy and user-friendly approach designed to make the specification of robotic swarm formations more accessible through digital drawings provided on a graphical user interface. Traditional approaches often require predefined information such as the exact number of robots and their desired pairwise distances, creating a boundary and limiting usability only to users within the controls and robotics field. To overcome those bounds, our method leverages computer vision techniques to interpret user-drawn sketches, efficiently extracting critical points that define the intended formation structure without needing any other user input. Our framework opens the door for non-expert users to deployrobot swarms, establishing desired robot formation. The extracted critical points, identified from user sketches, are utilized to systematically construct a rigid graph representation, accurately capturing the intended geometry and spatial relationships of the desired formation. To optimally assign individual robots to their respective positions within this formation, we integrate the Hungarian algorithm, renowned for effectively solving the linear assignment problem. This choice ensures an optimal, cost-effective allocation of robots to formation nodes, maintaining both spatial consistency and deployment efficiency. Furthermore, recognizing the importance of maintaining the shape of the desired formation, we conduct a theoretical analysis to ensure its structural integrity. This analysis verifies that the generated formation is free from unintended flips, thus preserving their intended configuration throughout deployment and operation. That analysis and guarantee are necessary since traditional formation control algorithms may allow for rotations, translations, and mirroring of specified formations, to happen; however, our approach requires user-defined formations that can’t tolerate potential flipping (i.e., mirroring) with respect to the original user-drawn sketch. Those occurrences take place when the drawings are non-symmetrical. Experimental and simulation results were conducted using the MATLAB Robotarium simulator and the UCI Robot Ecology Lab Robotarium, demonstrating the practical effectiveness and reliability of our proposed method. Results consistently indicate that formations generated via our approach match user-defined drawings with high accuracy and robustness. Moreover, the scalability and interactive nature of our solution highlight its suitability and adaptability for various complex multi-agent robotic coordination tasks across multiple domains, including search-and-rescue, environmental monitoring, agricultural automation, and collective transportation systems. This thesis, therefore, contributes to advancing human-robot interactions and enhancing multi-agent robotic operational efficiency.