Support: NSF grants IIS-0093581 and CCR-0330342
Abstract:
In this project we work on the drawbacks obtained in Balancing Sensing and Coverage in Mobile Sensor Networks: an Aggregation Based Approach. In such work, we assumed that each sensor was capable of correctly detecting and estimating every source that could appear in the region. Furthermore, the generalization of such work implicitly assumed a universal labelling for the sources. Here, we remove those assumptions, and present a novel min-max algorithm for the assignment. The algorithm minimizes the maximum penalty that can be imposed if one of the tasks is not fully attended. The limited sensing capabilities of each robot imply that there might be the need for agents that do not currently sense the source to help locate it. The issue is then how to assign sufficient number of robots to the sensing task and move them towards the source (even though some of them might initially not sense it) thereby diminishing the coverage of the region. We show that our algorithm converges towards a stable equilibrium point. The algorithm is shown to be optimal, fully distributed and thus scalable.Support: NSF grants IIS-0093581 and CCR-0330342
Abstract:
In distributed networks, each agents takes its decision based on the information it can gather by itself and the one it can share with its neighbors. Is thus required that the communication exchange between a particular pair of robots can be succesful. Unfortunately, communication systems are exposed to errors and hence there is some risk that the agents might work with incomplete information, leading the formation towards an incorrect state. Nonetheless, once motion is involved, the time constraints for communication exchange become very small for the network and the agents can afford to have several retransmissions in case of error. We quantify such intuition by obtaining bounds on the tolerance of error, the probability of exceeding such tolerance and the speed of the agents as function of the other variables.Support: NSF grants IIS-0093581 and CCR-0330342
Abstract:
In our previous works, we have been considering sensing and coverage as two essentially different tasks. This time, we approach the problem by considering sensing as a particular case of coverage, on which the density associated to the region has a high peak in a suitable neighborhood of the source location. The problem is then to define how the agents should interact with their neighbors, so all the agents in the formation agree on the global density function that should be assigned to the region.