OVERVIEW

S-MAC is a protocol designed specifically for the communication requirements and energy constraints of sensor neteworks.

Global schedule assignment: Support for multiple sleep schedules is a necessary feature for S-MAC in a large multi-hop network. However, nodes on the borders of multiple schedules consume more energy than others because they have to either wake up more often to follow different schedules or send more data on multiple schedules. Over time, these border nodes will exhaust their batteries more rapidly than other nodes, possibly causing network partitioning. To avoid this situation, we have designed algorithms that will incrementally switch all nodes onto a single global schedule across the whole network.

Even when migrating to a single global schedule algorithm, S-MAC still supports multiple schedules in the network. But we expect that they will only exist temporarily before they converge to the global schedule.  Multiple schedule support is still important for ad hoc and incremental network deployment.

One of the challenges in global scheduling is how to identify particular schedules. We have designed two algorithms to support global schedules. The first one is based on assigning a unique ID, such as a random number, to each schedule. Nodes will gradually follow the schedule with the smallest ID. Another algorithm is based on the age of each schedule. The one with the earliest birth time is chosen to be the global schedule.  Yuan Li implemented the first algorithm and demonstrated it on Mica2 motes at ACM SenSys'03. We are in the process of doing experiments in a large network to verify multiple schedules and the global schedule algorithms with different deployment patterns, and we expect to integrate this algorithm with our production code base in the coming quarter.

Link quality reporting: Working with researchers at Intel (Mark Yarvis) and UCLA (Eric Osterwile), we designed and implemented an API to export link quality information from S-MAC. S-MAC monitors packet losses and retransmissions from neighboring nodes, and passes the statistics to an independent link quality estimation component. This API is used by our emstar-based blacklisting module.

APPROACHES, ACCOMPLISHMENTS, AND FUTURE DIRECTIONS

Please look at MACSS project for details.

PEOPLE

FACULTY

Dr. John Heidemann

GRADUATE STUDENTS

Yuan Li
Wei Ye