Max Robotics — Global Service Robot Directory + US Certification

About

Richard Roberts is a leading roboticist whose work sits at the intersection of probabilistic mapping, visual odometry, and autonomous navigation in unstructured environments. His most influential contribution is the development of the Bayes Tree, an algorithmic foundation for probabilistic robot mapping that has garnered 97 citations and provides a rigorous framework for simultaneous localization and mapping (SLAM). Roberts pioneered memory-based learning for visual odometry, demonstrating how monocular cameras can estimate robot motion by learning mappings from sparse optical flow, bypassing traditional geometric calibration. His innovative two-part vision system for small robot navigation in forested environments combines saliency detection with model-based tracking, enabling quadrotors to incrementally build maps of tree trunks while estimating their trajectory. Roberts also advanced structure from motion through incremental light bundle adjustment and developed place recognition-based fixed-lag smoothing for environments with unreliable GPS. His work on optical flow templates for superpixel labeling represents a novel approach to obstacle avoidance, using learned general optical flow patterns to determine traversability. With over 200 total citations across his publications, Roberts has made foundational contributions to enabling robots to operate robustly in complex, GPS-denied outdoor environments.