Max Robotics — Global Service Robot Directory + US Certification

About

Martina Zambelli is a leading researcher in cognitive robotics and robot learning, whose work bridges the gap between autonomous systems and human-like interaction. Her primary research areas include cognitive architectures, deep reinforcement learning, and multimodal sensorimotor learning. Zambelli’s most influential contribution is the DAC-h3 cognitive architecture (74 citations), which enables humanoid robots to proactively explore and manipulate their environment through mixed-initiative interaction, grounded in biological theories of the mind. She also pioneered gentle object manipulation using curiosity-driven deep reinforcement learning (45 citations), allowing robots to handle fragile objects safely while minimizing wear and tear. Her work on online multimodal ensemble learning (35 citations) has advanced how robots develop self-learned sensorimotor representations for prediction and control. Zambelli has also contributed to multi-objective policy optimization (23 citations), addressing real-world trade-offs in robot learning. Notably, she is a co-author of the highly influential RoboCat (2023), a self-improving generalist agent for robotic manipulation that leverages heterogeneous robotic experience. Her research has been published in top venues like IEEE journals and conferences, and she continues to push boundaries with recent work on demonstration-led auto-curricula for sim-to-real transfer (DemoStart, 2025). With over 220 total citations, Zambelli’s work is shaping the future of adaptive, gentle, and socially-aware robots.