Flocking with formation control in mobile sensor networks for area search

Valbuena Reyes, Luis
Journal Title
Journal ISSN
Volume Title
University of Delaware
This thesis is addressing the integration of path following and formation control of a group of mobile robots to achieve flocking along specific path in the workspace. Several robots with similar capabilities are initially located in random positions and then move into a specified formation while performing path following over a designed route. The robots have to align their velocities to a common vector while approaching constant relative positions asymptotically. This problem has practical significance because it is the basic component for the development of a mobile sensor network that searches for and harvests information. Agent cooperation and coordination is motivated by several examples from nature where animals increase their chances of survival and gathering food by working in groups. Robot groups can apply the same idea to improve their coverage and reduce mission time. Wider coverage and versatility are some of the features that a (wireless) sensor network acquires when enabled to move, making it useful for surveillance and environmental inspection. Existing work has covered several aspects of the problem considered but it often ignores or tries to circumvent the discontinuities introduced into the system dynamics by the switching phenomena due to the operation of sensors and range of the communication. Another issue is the simplified dynamics (or kinematics) considered in cooperative control approaches, which raises some questions regarding its suitability for implementation on real platforms. Finally, although there is some limited work that combines formation control with flocking, these and other mission objectives -- such as path following -- may interfere with each other in existing formulations causing unpredictable behavior. To address these technical challenges we integrate various control designs into a coherent policy with guaranteed performance properties. The first of these components is a formation control via a nonsmooth artificial potential function that is in charge of making the robots fall into a formation specification while avoiding collisions among the robots. The second component is a flocking control law in concert with formation stabilization. A third component is path following to drive the formation along the designed route (does not require switching among the primitives that compose the route). To develop this methodology we had to make use of mathematical tools from nonlinear control and nonsmooth analysis that enable us to study both the convergence of our proposed control law over a nonlinear nonsmooth nonholonomic dynamic system.