Date of Award

Spring 4-2018

Embargo Period


Degree Type


Degree Name

Doctor of Philosophy (PhD)


Robotics Institute


George Kantor


The promised future filled with robots sensing and acting intelligently in the world is near fruition, thanks in part to continuous progress in robotic perception and state estimation. However, a number of challenges remain before state estimation systems and the robots that rely on them can be considered truly reliable. In particular, we must consider what happens when highly complex hardware and software systems designed and validated in laboratory environments enter the unbounded variety of reality. Will these systems fail innocuously or catastrophically? If so, how can we avoid or eliminate these failures to achieve reliable, robust behavior? The premise of this thesis is that engineering constraints and human finitude result in fallible systems that cannot compensate for all possible factors and situations. We refer to the collection of uncompensated factors as the con- text of a system, and propose that variations in context can explain why it is difficult to make state estimation reliable at scale. Vexingly, since context is, by nature, unknowable and unmodeled, we cannot rely on prediction and foresight to compensate for it. Instead, this thesis proposes that state estimation systems can adapt their behavior after deployment to the operating site to correct for unknown contextual effects. An example of this is the widespread and common practice of \parameter tuning", typically performed by a human expert to specialize a system to each deployment. To generalize this and other mechanisms of adaptation, we first develop a general theory of context in estimation and establish a statistical definition for estimation performance. We then develop a practical method for evaluating performance on-site without supervision, enabling estimation systems to observe the effects of context during operation. Finally, we explore automatic parameter tuning and experience-driven failure prediction as two methods of general adaptation. We demonstrate and validate this work on state estimation systems using online data from an instrumented automobile as well as online an indoor ground robot.