Max Robotics — Global Service Robot Directory + US Certification

About

Joshua Hoffman is a robotics and artificial intelligence researcher whose work sits at the intersection of multi-agent reinforcement learning and real-world multi-robot systems. His research focuses primarily on developing practical frameworks that enable teams of robots to coordinate effectively under decentralized control and partial observability — conditions that closely mirror the complexity of real-world deployment. Hoffman's most significant contribution is his pioneering work on macro-action-based deep multi-agent reinforcement learning, which addresses a critical gap in existing methods: the inability to handle asynchronous, variable-duration actions across cooperating robots. His 2020 paper introducing a centralized Q-Net for learning decentralized macro-action policies has garnered 27 citations, establishing it as a foundational reference in the field. By leveraging the Macro-Action Decentralized Partially Observable Markov Decision Process (MacDec-POMDP) framework, his research provides principled solutions to long-horizon coordination tasks that synchronized, primitive-action methods cannot practically solve. Across six published works spanning 2019 to 2025 and accumulating nearly 50 citations, Hoffman has consistently pushed the boundaries of scalable, decentralized robot learning. His body of work is particularly valuable for researchers and students seeking to bridge theoretical reinforcement learning with the messy asynchrony of real robotic systems.