Research

Current Projects
Control of Multi-Agent and Networked Systems
This project encompasses a range of issues involving not only autonomous agents but also human-in-the-loop systems. Although there have been considerable advances in control of autonomous robots in recent years, in practice, even a single unmanned vehicle typically requires a team of humans on the ground to monitor and control the mission. Our research is aimed at reversing this ratio so that a smaller number of humans can control a larger network of agents. Our research will contribute to the fundamental understanding of such networked robotic systems that must operate in complex, unstructured environments and communication over possibly unreliable communication networks with one another and over a distance with human operators. This research will integrate communication, control, and computing in ways that will both increase the level of autonomy possible in robot networks and the ability of humans and robots to work together in complex tasks. Robustness in such complex systems cannot be achieved by robustness guarantees for individual components alone but requires new solutions to many interdisciplinary research problems, such as stability and transparency in human-robot interaction, control over unreliable communication networks, and sensor-based control of multi-agent systems.

Geometric Methods in the Control of Bipedal Walking Robots
The object of this project is to develop control algorithms for bipedal walking robots using concepts of geometric reduction. Geometric reduction takes advantage of certain symmetry properties in the dynamic equations of bipedal robots and facilitates the computation of lower dimensional models for analysis and control design. The goal is to enable control concepts developed for two-dimensional, or planar, bipeds to be applied to fully three-dimensional bipeds. In this way, the complexity of the control problem for three-dimensional walking robots is greatly reduced. The project will also investigate the effects of asymmetries in leg parameters, such as leg length and leg mass, on the existence and properties of passive gaits. Preliminary investigations indicate that asymmetry in leg parameters, such as leg mass, results in qualitative changes in gait, such as period-doubling bifurcations and chaotic motion. Deliverables from this research include new control algorithms, new simulation models and graphical simulation tools for visualization of simulation data.

Reliable and Robust Control of Formations of Unmanned Vehicles
This project is to develop reliable and robust control architectures and algorithms for networks of autonomous aerial and ground vehicles. The aim is to develop control laws that have low sensitivity to noisy and lossy data communication among vehicles, that are scalable in terms of number of vehicles, and that have the ability to handle discrete transitions in the network, such as formation reconfiguration, addition or loss of vehicles from the formation, etc. Applications of this work include undersea and planetary exploration, search and rescue, air traffic control, and control of sensor networks. Both theoretical and experimental issues are being investigated.