Columbia University

Controlled Mobility

  • OoueiyvBfNWNC0TWpOqNN1-0lJudBofPkpW4BS9RSfI
  • research-controlled-mobility-01

Mobility gives rise to fundamentally new problems that involve the interplay between various system functions. For instance, the mobility of autonomous vehicles is controlled through communications while at the same time the ability of the network to communicate is greatly affected by mobility. We studied a hierarchical wireless networking approach where some of the nodes which are more capable than others can serve as mobile backbone nodes and provide a backbone over which end-to-end communication can take place. Our approach consists of controlling the mobility of the backbone nodes in order to maintain connectivity. We developed distributed approximation algorithms for minimizing the number of backbone nodes that perform well under mobility.

Most of controlled mobility algorithms have been studied in the past only analytically and via simulation. Therefore, in collaboration with the Distributed Network Analysis Group, we established a mobile networking testbed composed of mobile iRobot Creates equipped with IEEE 802.11 access points. We used the testbed to evaluate the performance of the Spreadable Connected Autonomic Network (SCAN) algorithm which is a fully distributed, online, low overhead mechanism for maintaining the connectivity of a mobile network. To the best of our knowledge, this was the first evaluation of such an algorithm in a mobile testbed and it allowed us to obtain important insights regarding the tradeoffs between various system parameters.

In the video below, Joshua Reich demonstrates the operation of the testbed composed of several iRobot Creates that was described in the following paper.

Finally, we studied changes in the dynamic graph structure that represents a mobile wireless network that evolves over time due to node mobility. Quantifying the changes in the graph structure is important for understanding the behavior of higher-layer network algorithms. We defined several graph evolution metrics and evaluated them through extensive numerical simulations under Levy Walk mobility. We considered the effects of the rate of graph change on the performance of network protocols and showed that the proposed metrics are viable for quantitatively measuring the change in a sequence of evolving graphs.