Home
About
The Intelligent Motion Lab studies motion planning and control for intelligent robotic and biological systems. Our research applies computational, mathematical, and statistical techniques to address the complexity of high-dimensional motion in unstructured, dynamic, and uncertain environments, and seeks applications to domestic and industrial robotic manipulation, human-robot interaction, robot- and computer-assisted surgery, and legged locomotion. We are also interested in using computational theories to inform the study of motor cognition in humans and other biological systems. The lab is directed by Prof. Kris Hauser and is part of the Indiana University School of Informatics and Computing.
IML bridges theory to practice by combining high-performance computing and simulation tools with a variety of robot hardware. Our overall goal is to contribute new tools, techniques, and theories that will help machines perform sophisticated physical tasks with high value to society.
Research Areas
For more information, please see the research page.
Integrating Task, Contact, and Motion Planning
In order for robots to perform increasly sophisticated manipulation tasks in the home and in industrial settings, they must be able to coordinate the motion of many joints to achieve symbolic tasks, involving intricate changes of contact. Similar problems are faced by legged robots traversing rough terrain. We aim to build new tools and theory for integrating discrete task planning (typically considered a domain of AI) and continuous motion planning (considered a domain of robotics) within a single framework. Our planners have addressed manipulation and legged locomotion problems in configuration spaces with up to 42 dimensions, and we have proven the asymptotic reliability properties of our algorithms in general settings. Current work is studying programming support for rapid planner prototyping, as well as general-purpose planning strategies that exploit problem structure and learn from past experience.
Real-Time Planning in Uncertain and Dynamic Environments
Dynamic, unpredictable environments require an intelligent agent to reactively adapt to incoming sensor input and, often, to proactively seek new information. We are studying new real-time architectures to achieve goal-oriented decision-making while dealing with unpredictable disturbances and partial observability. We are applying a replanning based approach to robots that have limited sensors, that have limited capacity to represent their environment, and that interact with unpredictable agents. As opposed to hierarchical architectures that separate path planning from low-level control, replanning techniques can exploit advances in computing power (including parallel and stream computing) in order to produce responsive, convergent, safe behavior even under complex dynamic and environmental constraints. Our recent results include an adaptive time stepping technique that addresses the problem of high variance in planning time to yield provably safe and complete execution in real-time.
Shared Human-Robot Decision Making
Semi-autonomous robots have the potential to combine the adaptability, contextual awareness, and intuition of humans with the precision, availability, and consistency of robots. We are interested in studying tightly coupled, real-time human and robot decision-making, where high performance of the overall system requires synthesizing both agents’ strengths in cognition and sensing. We are considering applications to puzzle solving, active safety systems for automobiles, interactive CAD systems, and industrial inspection and material handling.
