Rate My Professor Joni Pajarinen

Pajarinen's research specializes in robot learning, multi-agent reinforcement learning, imitation learning, mobile manipulation, decision making under partial observability, and autonomous systems. His work enables robots to learn independently, operate in unknown environments, and assist humans proactively, with applications in heavy machinery, underwater vehicles, and humanoid robots. As principal investigator, he directs projects including ENGAGE for training engineers in mobile working machines, XSCAVE for explainable safe control in heavy machinery, Aurora, IWM/Tuomisto/T41020, and MARL for efficient multi-agent reinforcement learning. Key publications encompass An Algorithmic Perspective on Imitation Learning (2018, Foundations and Trends in Robotics), Partially Observable Markov Decision Processes in Robotics: A Survey (2022, IEEE Transactions on Robotics), Self-Paced Deep Reinforcement Learning (2020, NeurIPS), Robotic Manipulation of Multiple Objects as a POMDP (2017, Artificial Intelligence), Curriculum Reinforcement Learning via Constrained Optimal Transport (2022, ICML), and AgentMixer: Multi-Agent Correlated Policy Factorization (2025, AAAI). With 85 research outputs, including 45 conference papers and 25 journal articles, Pajarinen advances AI-driven sustainability, productivity, and safety in robotics. He has presented at conferences such as ICML 2008 and MLSP 2008 and held positions of trust, including as Chair at a university external organization in 2011.