Research Seminars @ Illinois

Abstract: Robots' sensors provide information that is seldom, alas, a perfect characterization of state. This challenge has been tackled through classic techniques such as probabilistic estimation and filtering, as well as uncertainty-aware planning. Such approaches are effective when uncertainty may be appropriately modeled as noise (e.g., degradation or corruption via randomness). Instead, this talk examines a rather different species of non-ideal observability: we examine circumstances in which sensing information is sporadic, with high-quality data arriving only intermittently. The challenges then include temporal sparsity, with the robot having to act with estimates that are out-of-date. After presenting our bounding-based approach to the problem of planning for such settings, I will explain some counter-intuitive examples (e.g., more frequent information isn't always better).  A core underlying idea is that meta-knowledge (i.e., knowledge about knowledge) plays a specific role for the robot, and in this case such meta-knowledge describes when future information will arrive, but not what that information will be. I will then discuss the question of how and when to establish such knowledge, including via negotiation between cooperating parties.  It will turn out that part of this involves robots planning to re-plan.

Bio: Dylan Shell is a computer scientist at Texas A&M University who works in the areas of robotics and AI. Broadly, his research aims to analyze and synthesize complex, intelligent behavior in systems that interact with the physical world.  He has an interest in algorithmic and formal foundations of planning problems, and extremely simple (or minimal) robots. He has published papers on topics from multi-robot task allocation, biologically inspired multiple robot systems, estimation of group-level swarm properties, rigid-body simulation and contact models, and robotic theatre.  His work has been funded by DARPA, DoD (ONR, ARL), NSF as well as Ford and 3M; he has been the recipient of teaching, service and reviewing, and research awards.  Dylan serves as the President of the Robotics Science and Systems Foundation.