These are some of the projects topics I'm involved in.
For more information on current research, please refer to the latest publications.
Efficient coordination of multi-vehicle sensor systems
The use of groups coordinated vehicles is revolutionizing the way in
which observation, measurement, and estimation is carried out in
aerial, underwater, and ground environments. In particular,
multi-vehicle sensor systems are uniquely suited for active monitoring
and coordinated action in remote areas. These versatile systems are
envisioned to perform a variety of spatially-distributed tasks such as
search and recovery, data collection and fusion, localization of
chemical pollutants and reconnaissance.
In order to develop cooperative algorithms for the efficient
coordination of multi-vehicle systems, one should consider the
different capabilities that individual vehicles possess. For example,
the cooperative strategies should take account the possible vehicles'
energy constrained operation, their sensing, communication and
computation capabilities, and other dynamical restrictions. The
following simulation shows the final emerging behavior that results
from implementing a simple local rule for a group of robotic agents
with a limited sensing radius. The final configuration is locally
optimal for coverage and, if the sensing radiuses are large enough,
coverage of the dark blue areas will be guaranteed.
Modeling and control of robotic networks
Cooperative robotic networks are a paradigm of networked systems. Their study already poses many of the questions involved with general networked systems. For example it is necessary to analyse the different trade-offs in communication, mobility, computation and sensing to produce energy-efficient and truly scalable algorithms. To address these issues we have proposed a preliminary model for cooperative robotic networks that is helpful in quantifying the performance of different distributed algorithms. An example of a distributed algorithm that we have completely characterized is the agree (on direction) and pursuit algorithm simulated below that is related to a "leader-election" algorithm.
Based on different priorities, the agents change their direction (and color) as soon as the meet another agent with higher priority going in a different direction. One can see how all agents end up going in the same direction (have same color). While the choice of direction occurs relatively fast (in O(n) time) through local interactions only, the second objective of achieve a uniform distribution over the circle is only achieved asymptotically (in O(n^2 log n))
Motion coordination for adaptive sampling in multivehicle systems
A main objective of this project is the development of scalable, systematic approaches to the synthesis of adaptive sampling strategies by mobile sensor networks. In particular, the research challenges have been identified with a specific application in mind: the study of atmospheric aerosol-dust-cloud-radiation interactions, and how aerosols and dust affect cloud microphysics and, more generally, climate dynamics. A multi-vehicle test bed that will carry out the projected sampling is being developed at SIO and consists of a group of several UAVs. The successful completion of the project requires progress on the development of distributed data fusion algorithms that help estimate spatio-temporal processes and coordination algorithms that exploit these environmental filters.
Geometric control of mechanical systems
We are interested in the analysis and design of single underactuated mechanical systems. Underactuated systems possess fewer actuators than degrees-of-freedom. The interest in underactuated systems stems from two considerations: first, underactuation allows software control redundancy in the case that one or more actuators of a mechanical system system fails; second, it is possible to replace expensive mechanical elements with advanced control algorithms which are inexpensive to compute. Underactuation can be found in all sorts of autonomous vehicle systems and other biomimetic robots like "walkers", snakes or fishes. The analysis and design of these type of systems requires specialized tools such as geometric and nonlinear control techniques, that account for the special models capturing their dyanamics and help produce low-complexity representations for their control.