Andrea Goldsmith

Papers

Total Citations

H-Index

About

Andrea Goldsmith is a pioneering researcher in multi-robot systems and adversarial machine learning, whose work focuses on creating resilient, trustworthy AI for collaborative teams. Her major contributions lie at the intersection of decentralized decision-making and security, where she develops algorithms that enable robots to cooperate effectively even when some agents are malicious or untrustworthy. In her most-cited work (2023, 14 citations), Goldsmith introduced a novel framework for resilient hypothesis testing in multi-robot crowdsensing tasks, exploiting stochastic trust observations between robots to make tractable, secure decisions at a centralized fusion center. This breakthrough addresses critical vulnerabilities in real-world deployments, where adversarial robots could otherwise compromise mission outcomes. Her subsequent research extends these principles to decentralized cooperative bandit teams (2022, 3 citations), where she designs partner-aware algorithms that enable AI agents to consider team-wide consequences rather than acting greedily—mimicking human collaborative intuition. By bridging trust theory, game theory, and multi-agent reinforcement learning, Goldsmith is shaping how autonomous systems can safely operate in contested environments, from disaster response to defense applications. Her work is essential reading for anyone interested in building AI that is both intelligent and trustworthy.