Max Robotics — Global Service Robot Directory + US Certification

About

Troi Williams is a roboticist whose research centers on state estimation, sensor modeling, and active perception for autonomous systems operating under severe constraints. Williams’s major contributions lie in developing methods that enable robots to localize effectively when sensor data is limited, costly, or unreliable. In their most cited work (5 citations), they introduced a two-stage transfer learning approach for learning state-dependent sensor measurement models with minimal ground truth data—a critical advance for robots that must learn many sensor models without exhaustive data collection. Williams also pioneered DyFOS, an active perception framework that dynamically selects optimal sensor states to minimize localization uncertainty while navigating occlusions and obstacles, particularly for perception-denied rovers reliant on a viewer robot. Their POMDP-based approach to deciding *when* to localize addresses the practical challenge of resource-constrained robots, such as submersibles that must surface to obtain position fixes. By tackling the fundamental tension between accurate localization and operational efficiency, Williams’s work has direct implications for field robotics in GPS-denied environments, planetary exploration, and underwater autonomy. Their research bridges theoretical rigor with real-world deployment constraints, making them a rising voice in practical state estimation.