- This event has passed.
Foundations of Resilient Collaborative Autonomy: From Combinatorial Optimization to Control and Learning
March 4, 2020 @ 12:00 pm - 1:00 pm
Vasileios Tzoumas (MIT)
E18-304
Event Navigation
Abstract:
Collaborative autonomous robots promise to revolutionize transportation, disaster response, and environmental monitoring. Already, micro-aerial vehicles have become a multi-billion-dollar industry; and in this new decade, teams of semi-autonomous ships, cars, and underwater exploration vehicles are being launched. A future of ubiquitous autonomy is becoming a reality, where robots can autonomously split into teams, and navigate and learn the world. However, this future is threatened by attacks and failures that can compromise the robots’ teams, control, and learning capabilities, by either deceiving the robots or (temporarily) disabling them. These threats lie outside the reach of cybersecurity and classical control. To counter them, novel algorithms are needed at the intersection of control, learning, and combinatorial optimization that can balance complementarity with redundancy, and efficiency with robustness. I will present two algorithms from my research, as well as my vision for a resilient collaborative autonomy in dynamic, resource-constrained environments.
First, I will present the first near-optimal algorithms for robust combinatorial optimization against any number of robot-disabling attacks. The algorithms robustify for the first time robotic teams and their plans against such attacks. Besides the theoretical guarantees, I will demonstrate this via experiments of multi-robot information gathering for search and rescue, and target monitoring. Second, I will present algorithms that enhance robotic perception against deceptive failures. The algorithms achieve extreme robustness in near real-time for the first time. I will illustrate this across datasets for localization and mapping (SLAM), object recognition, and 3D reconstruction. All algorithms above apply beyond multi-robot systems: from facility location problems in social and supply networks to robust estimation in power networks. I will conclude with my vision for a collaborative autonomy that is not only robust but also resilient: I will argue the need for a technological convergence between (i) “cyber” capabilities for a distributed artificial intelligence, driven by adaptive learning and resource-aware robust control algorithms, and (ii) “physical” capabilities of self-reconfigurable systems, self-healing materials, and smart devices.
Bio:
Vasileios Tzoumas is a research scientist at the Department of Aeronautics and Astronautics (AeroAstro), and the Laboratory for Information and Decision Systems (LIDS), Massachusetts Institute of Technology (MIT). Before that, he was a post-doctoral associate at AeroAstro and LIDS for a bit over a year. He received his Ph.D. in 2018 at the Department of Electrical and Systems Engineering, University of Pennsylvania (UPenn). In 2017, he was a visiting Ph.D. student at the Institute for Data, Systems, and Society, MIT. He holds a Diploma in Electrical and Computer Engineering from the National Technical University of Athens (2012); a Master of Science in Electrical Engineering from UPenn (2016); and a Master of Arts in Statistics from the Wharton School of Business at UPenn (2016). He works to enable autonomous, collaborative cyber-physical systems that are resilient against denial-of-service and deceptive attacks and failures. His theoretical focus is at the interplay of resource-aware control, learning, communication, and computing. His application and experimental focus include multi-robot tasks of autonomous navigation for surveillance and information gathering. Vasileios builds on fundamental tools of control theory, computational complexity, robotic perception, game theory, and non-convex and combinatorial optimization. He was a Best Student Paper Award finalist at the 2017 IEEE Conference in Decision and Control (CDC).