Max Robotics — Global Service Robot Directory + US Certification

About

Anna Helena Reali Costa is a prominent Brazilian researcher whose work spans reinforcement learning, multiagent systems, autonomous robotics, and knowledge transfer in artificial intelligence. Based at the University of São Paulo, she has made substantial contributions to accelerating machine learning through the intelligent incorporation of prior knowledge, heuristics, and human guidance. Her most influential work, "Simultaneously Learning and Advising in Multiagent Reinforcement Learning" (2017, 104 citations), addresses the fundamental challenge of sample inefficiency in multi-agent settings by enabling agents to both share and refine advice during learning. This theme of knowledge-guided acceleration runs throughout her career, from her early research on heuristic action selection (79 citations) to stochastic abstract policies that generalize learned knowledge across tasks (45 citations). Her contributions to qualitative case-based reasoning further demonstrate a commitment to bridging symbolic and statistical AI approaches. Beyond learning algorithms, Costa has tackled practical robotics challenges, including cooperative multi-robot localization, vision-based systems for humanoid soccer robots, and object-based SLAM, reflecting a deep engagement with embodied AI. Her early inverse reinforcement learning work (2006) also places her among pioneering researchers in learning from human intent. With a citation profile exceeding 340 citations, her interdisciplinary impact continues to shape both theoretical and applied AI research.