SECURE MULTI-ROBOT COMPUTATION FOR HETEROGENEOUS TEAMS : FOUNDATIONS AND APPLICATIONS

Abstract

In the rapidly evolving field of robotics, significant progress has been made in the planning, control, and coordination of multi-robot systems, embedding robots into various sectors such as household, manufacturing, healthcare, and surveillance. Despite these advancements, challenges arise, particularly concerning privacy due to robots' potential to access and share more information than necessary, risking sensitive data exposure. Addressing this, our research introduces innovative strategies to ensure collaborative computation among robots while safeguarding privacy, thereby preventing unnecessary information sharing and achieving optimal objectives. We propose lightweight communication protocols for data synchronization, reducing the need for extensive data exchange, and a secure multiparty auction-based algorithm for private task allocation without revealing sensitive data. Additionally, we explore the use of secure multiparty computation with Markov Decision Processes (MDP) for planning, ensuring privacy in multi-agent cooperation. Building on this foundation, we delve into decentralized multi-robot information gathering (DMRIG), presenting the Asynchronous Information Gathering with Bayesian Optimization (AsyncIGBO) and Distributed and Decentralized Robotic Information Gathering (DDRIG) algorithms to improve environmental monitoring data collection efficiency, balancing communication complexity, and privacy. Through practical experimentation, these algorithms' real-world efficacy is demonstrated, emphasizing their role in enhancing environmental monitoring via sophisticated information sharing and task allocation among robots. This dissertation provides a comprehensive approach to addressing privacy and efficiency in heterogeneous robot systems, showcasing the potential of these technologies to advance robotics applications securely and effectively. Together, these components form a comprehensive approach to addressing privacy concerns in heterogeneous robot systems. By interlinking efficient data sharing protocols, secure task allocation, private planning strategies, and optimized multi-robot information gathering, the dissertation lays the groundwork for a new paradigm in robotic collaboration. This synergy ensures that robots can work together effectively, achieving optimal objectives without compromising sensitive information, marking a significant advancement in the field of robotics.