News & Events

Navigating the Nexus of Conflict and Cooperation: Strategic Planning in Autonomous Multi-Agent Systems amidst Adversaries

Bhaskar Vundurthy, PhD, Project Scientist at the Robotics Institute, Carnegie Mellon University

1400 Wegmans Hall

The increasing presence of adversarial robots in complex missions, such as disaster response, surveillance,  or reconnaissance, necessitates the development  of strategic planning algorithms for heterogeneous multi-robot teams. To prevail in these battlespaces, such algorithms must promote cooperation within robotic teams, thrive amidst adversarial interference, and capitalize on information asymmetries wherever possible. Achieving these capabilities requires breaking down disciplinary silos and fostering collaboration across various fields. In this talk, I will introduce an interdisciplinary framework that seamlessly blends concepts from computational geometry, game theory, control theory, and operations research to guide this research along three primary thrusts: 1) Collaborative Autonomy that optimizes coordination within the same team, 2) Contested Coordination that develops retaliation algorithms to evade or intercept adversaries, and 3) Informed Decision-making that leverages intelligence disparities to exploit adversary vulnerabilities. Grounding this research in experimental validation, this talk will demonstrate the practical feasibility and adaptability of these algorithms on in-house fabricated heterogeneous multi-robot testbeds, highlighting how they address real-world challenges in the field. Finally, I will illustrate this framework with a multi-robot task allocation example that showcases collaboration on shared tasks, handles competing objectives, and leverages information distribution to minimize ergodicity. Looking ahead, this framework lays the groundwork for advancements in adversarial multi-robot coordination, enabling the development of increasingly sophisticated and autonomous robotic systems.

Dr. Bhaskar Vundurthy is a robotics researcher specializing in adversarial heterogeneous multi-agent systems. As a project scientist at the Robotics Institute, Carnegie Mellon University (CMU), he investigates the complex interplay between collaborative agents and adversaries in dynamic environments, developing strategies to ensure effective teamwork, robust performance, and resilience against attacks and sabotage. He actively utilizes a variety of robotic platforms to validate and refine his research findings. This work has led to publications in premier IEEE journals and conferences, with both best-paper nominations and special-issue journal invitations. During his time at The MathWorks Inc., Bhaskar created advanced examples for autonomous driving in simulated environments, specifically for the lndy-500 oval circuit.

Bhaskar thrives on mentoring and empowering students to excel in research. He has mentored over 20 graduate students, many of whom have gone on to publish high-impact papers, secure competitive fellowships, and pursue careers in academia and research. His dedication to student success has been recognized through a nomination for CMU's prestigious 'Andy Award' and winning a 'Best Teaching Assistant Award.' Beyond his academic pursuits, Bhaskar finds balance and inspiration in gardening and nature photography.