Naomi Ehrich Leonard, Mechanical and Aerospace Engineering, Princeton University
This project addresses critical, open problems in using feedback to control the shape and geometry of a vehicle formation and enable high performance use of vehicle groups. The strongest motivation comes from using sensor-equipped vehicles as a mobile sensor network. In this context group shape plays a critical role; e.g., it should be adapted to minimize error in estimates of gradients or boundaries in a sampled field. The main objective of this project is to develop a methodology that systematizes the design and analysis of control over the shape of a vehicle formation using interconnection that derives from models of tensegrity structures. Mobile, multi-agent systems can have complex, multi-scale dynamics; the need to develop systematic and scalable control design for such systems requires new approaches that go beyond existing theory and algorithms. Tensegrity structures are spatial networks of interconnected struts, cables and rods that have remarkable stabililty and rigidity properties. The proposed approach will derive vehicle control laws that mimic forces internal to tensegrity models to produce spatial networks of vehicles that behave like tensegrity structures and, in particular, inherit the same stability properties. Tools from mechanics can be applied to study the controlled vehicle formation, since the controlled dynamics are those of a mechanical system. The first goal is to solve the "reverse engineering" problem: given an arbitrary shape in 2D or 3D, a method will be derived, using a modification of existing tensegrity models, to define and prove dynamic stability of a tensegrity with this shape. Next, the method will be extended to smoothly control the formation from one shape to another, using a path that consists of tensegrity structures. Shape changes will be integrated with motion control algorithms. Low-level controllers for trajectory tracking will be considered as will coordination and stable control of rigid body dynamics.
Groups of vehicles, equipped with sensors to measure the environment, have enormous potential to revolutionize the way that monitoring, estimation, detection and learning can be performed in the air, on land and in the water. With well choreographed and coordinated motion of the sensing vehicles in the group, the measured data can be made maximally information rich and can have the greatest impact on the issue at hand. Monitoring forests for fire and croplands for damage from the air, tracking phytoplankton blooms or endangered whale pods in the ocean are just some examples. The PI leads a team of oceanographers and engineers in an effort to develop a sustainable ocean observing and prediction system using a coordinated network of sensor-equipped, autonomous underwater vehicles. In this and other applications, the technology has already begun to contribute to improved understanding of ecosystems and the global climate, prediction of safe conditions in coastal environments to deploy relief boats, improved methods for detecting and tracking chemical plumes, spills and red tides, new means for search and rescue and more. Control and adaptation of the shape, geometry and pattern of the moving vehicle formation play a critical role in optimizing performance in monitoring, estimation, detection and learning. This project focuses on design of systematic and reliable algorithms for control and adaptation of the shape, geometry and pattern of a vehicle collective. The research promises significant impact on a wide range of issues of national interest from security to the environment.
