Research Areas
Scalable Control with Compositional, Hierarchical, and Learning-based Approaches
|
Existing control design and verification methods are severely limited in their applicability to large scale
systems, such as infrastructure networks and coordinated fleets of autonomous vehicles. The growing
sophistication of these systems demand equally sophisticated control methods that are applicable to the
complex requirements and large-scale dynamical models describing their operation. To develop scalable control
design and verification methods that address large systems with complex requirements, we are pursuing
compositional, hierarchical, and learning-based approaches. The aim in compositional analysis is to derive
system-level guarantees from properties of the subsystems, thus breaking apart intractably large problems into
subproblems of manageable size. The aim in hierarchical control is to decompose the synthesis and verification
tasks into layers, from high-level decision making to low-level control synthesis. As an example, the control
of an autonomous vehicle may include a routing layer at the top that selects a sequence of road segments,
which is then broken into maneuvers at the lower layer, followed by an online motion planner that provides
reference signals for a low-level tracking controller. We are also developing data-driven methods for control
design and verification, and using statistical learning theory to provide probabilistic performance
guarantees and associated sample complexity bounds. The results of the theoretical research are applied to the
research areas below.
|
Traffic Management for Smart Cities
|
Connected vehicles and infrastructure, combined with increased automation, promise to minimize congestion,
enhance safety, and lower emissions. We are pursuing a hierarchical traffic control framework that integrates
feedback strategies at the vehicle, road link, and network layers. At the vehicle layer we investigate how
vehicle fleets (“platoons”) can dramatically increase the capacity of urban roads and intersections by
maintaining a constant spacing between them (see this video for an
experimental demonstration). At the road link layer, we bring tools from formal methods to develop reactive
signal timing policies for finite horizon objectives, such as dissipating queues and avoiding saturation. At
the network layer we study routing games, particularly focusing on transient phenomena rather than assuming
equilibrium conditions. A particular focus of this work is to enable better response systems to assist local
authorities in managing extreme demand, such as when entire counties have to be evacuated to protect the
residents from a wildfire. In a separate project we are studying real-time allocation of curbspace, which is
increasingly becoming a dynamic interface between people and vehicles instead of stationary parking space. We
are developing pattern recognition tools using camera images to identify curb usage patterns, with the aim of
dynamic allocation to different types of activity.
|
Spatial-Temporal Patterning in Biological Systems
|
As another application of the compositional and hierarchical approaches described in the first area above, we
are investigating how spatial and temporal patterns emerge from local interactions in biological systems, both
natural and synthetic. Such biological systems are typically high dimensional and nonlinear, involving
interactions among neighboring cells as well as between cells and the environment, as in the emergence of gene
expression patterns in developing embryos. Our research reveals theoretical conditions for spontaneous pattern
emergence, as well as new perspectives for understanding how initial patterns are refined into later patterns.
In recent work, we collaborated with experimental biologists to empirically demonstrate spontaneous
contrasting patterning in systems of genetically engineered microorganisms, including colonies of bacteria
communicating through diffusible molecules, as well as engineered yeast cells interacting by way of a
computer-controlled light input. Our research provides insight into natural biological mechanisms and lays the
groundwork for theory-based design of synthetic biological systems that can be used for microbial consortia,
biomaterials, and tissue engineering. A further objective of this research area is to help synthetic biology
develop into a scalable engineering discipline capable of creating complex, autonomous and controllable
multicellular behavior.
|
|