Toward controllable morphogenesis in large robot swarms

Carrillo-Zapata, Daniel; Sharpe, James; Winfield, Alan Frank T.; Giuggioli, Luca; Hauert, Sabine

Abstract

Morphogenetic engineering aims to achieve functional, self-organized but controllable structures in human-designed systems. Controlling the structures is crucial if they are to be used for real-world applications. Building on previous work on morphogenesis, in this letter, we present a new algorithm, with controllability at its core, for large swarms of simple robots where morphogenesis occurs without self-localization, predefined map, or preprogrammed robots. Controllability is achieved through three parameters that influence the morphogenesis process and create a rich morphospace of quantitatively different shapes. The algorithm was tested in over 2000 simulations and three times on real swarms of 300 kilobots. Swarms were able to grow shapes using only local communication, and regrow missing parts when manually damaged. Extra simulations also demonstrated swarms adapting to an obstacle in the environment by getting around it. Results were compared with our previous work on morphogenesis to show how controllability allowed richer shapes. This letter represents a step into designing a controllable morphogenesis algorithm toward more functional swarms for real-world applications.

• Journal Article Type: Article
• Acceptance Date: Jun 19, 2019
• Online Publication Date: Jul 4, 2019
• Publication Date: Oct 1, 2019
• Deposit Date: Oct 23, 2019
• Journal: IEEE Robotics and Automation Letters
• Print ISSN: 2377-3766
• Publisher: Institute of Electrical and Electronics Engineers
• Peer Reviewed: Peer Reviewed
• Volume: 4
• Issue: 4
• Pages: 3386-3393
• DOI: https://doi.org/10.1109/LRA.2019.2926961
• Keywords: Artificial Intelligence; Computer Vision and Pattern Recognition; Computer Science Applications
• Public URL: https://uwe-repository.worktribe.com/output/4041373
• Publisher URL: https://doi.org/10.1109/lra.2019.2926961