Support:
NSF grants IIS-0093581 and CCR-0330342
How to get a network of robots to organize themselves to that we can:
without compromising such coverage?
A first suboptimal approach is presented, on which it is guaranteed that at least one agent would be doing sensing, and at least one agent would be doing coverage.
Although it can be guaranteed that such algorithm leads the formation towards a set of stable equilibrium points on which each task has at least one associated agent, nothing can be said about its scalability properties.
An alternative more robust approach considering Lyapunov functions for each individual task is studied.
A hybrid control rule is defined by comparing the value of the associated Lyapunov functions, which can be evaluated locally by each robot in the formation.
Such comparisson might induce a sliding mode among the agents, preventing them to complete any of the tasks but guaranteeing that the sacrifice in each one of them is the optimal sacrifice.
The system is shown to evolve towards an equilibrium position defined by agents of three types:
Agents completing their local sensing task.
Agents completing their local coverage task.
Sliding agents which find a compromise between the two tasks in the network.
In order to avoid the sliding mode, a constant correction term is added to the initial Lyapunov functions.
Such correction term is defined beforehand, assuming some knowledge on the size of the network.
Note: The simulations on this project have been based on the Mathematica code provided by the authors of Coverage Control for Mobile Sensing Networks. IEEE Transactions on Robotics and Automation, 20(2):243--255, 2004.
The slides of the talk given in the 2006 IEEE International Conference in Networking, Sensing and Control are available here.