1. ct-ee@eecs UC Berkeley, EECS Congestion Control and Fairness for Many-to-One Routing in Sensor Networks Background  sensor motes route data back to base station  routing structure is that of a tree, rooted at base station Problem  hard to estimate optimal sending rate  with large network, can easily congest motes closer to the base station  also, base station typically receives more packets from motes nearer than from motes further away Cheng Tien Ee

2. ct-ee@eecs UC Berkeley, EECS Congestion control solution  each mote locally measures transmission rate  divides that rate by the total number of downstream motes  disseminate that information downstream Fairness implementation  we say it is fair if same number of packets received from each mote over a period of time  each mote maintains per child FIFO queue  in each transmission epoch, send same number of packets from each queue as number of downstream motes for that queue fairness implementation results: 10 mote network red line: our algorithm, EPS green line: round-robin

3. ct-ee@eecs UC Berkeley, EECS Conclusion network adapts itself automatically  impossible to manually configure 1000s of motes exact, same, simple code runs in each mote  reduces programming, debugging time very scalable  size of queues can be small, constant  state required increases linearly with number of neighbors congestion control and fairness can be implemented in transport layer, thus can be used with different MAC protocols  little or no modifications to MAC