Max Robotics — Global Service Robot Directory + US Certification

About

Joshua Song is a leading researcher in autonomous robotics, with a primary focus on decision-making under uncertainty and robust manipulation. His most significant contributions center on advancing the practical application of Partially Observable Markov Decision Processes (POMDPs) for real-world robots. Song’s seminal work, the TAPIR software toolkit (40 citations), provides an online, anytime framework for approximating POMDP solutions, directly addressing the computational intractability that has long hindered their deployment. This foundational contribution enables robots to compute motion strategies reliably despite sensor noise and control errors. Demonstrating the real-world viability of his theories, Song led the "POMDP-Based Candy Server" project, a landmark seven-day public demo that showcased a robot autonomously making long-term decisions in a dynamic environment. Further expanding robotic capabilities, he pioneered the exploitation of trademark databases for object recognition, a novel approach that allows service robots to identify and fetch household items without tedious manual data collection. His distributed, any-time robot architecture, validated on the MOVO mobile manipulator, ensures robust grasping and manipulation. Through these contributions, Song has bridged the gap between complex theoretical planning frameworks and practical, deployable robotic systems.