1. Adaptation of TDMA Parameters Based on Network Conditions Bora Karaoglu Tolga Numanoglu Wendi Heinzelman Department of Electrical and Computer Engineering University of Rochester, NY, USA Bora Karaoglu Tolga Numanoglu Wendi Heinzelman Department of Electrical and Computer Engineering University of Rochester, NY, USA

2. Motivation  Capacity  Each tx occupies some part of the capacity

3. Motivation  Clustering approach:  Divide into a number of chunks  CHs use chunks  Question?  How many chunks?  Work summarized in:  Analytical model  Optimization

4. Agenda  Protocol Overview: MH-TRACE  Analytical Model  Dropped Packets  Collisions  Proof of Concept  Optimization of TDMA parameters

5. Multi-Hop Time Reservation Using Adaptive Control for Energy Efficiency  TDMA  Soft clustering  CHs responsible for channel access only  Inter cluster communication is allowed  TDMA  Soft clustering  CHs responsible for channel access only  Inter cluster communication is allowed N f = 6

6. Protocol Overview: MH-TRACE  Factors limiting performance:  Dropped Packets  Real-time communication  Limited Local Capacity  Clustering  Uneven distribution of Load  Node Distributions  Mobility  Collisions  Spatial Reuse  Limited capacity Divisions

7. Agenda  Protocol Overview  Analytical Model  Dropped Packets  Collisions  Proof of Concept  Optimization of TDMA parameters

8. Analytical Model  Shortcomings of Simulations  Substantial Processing Power and Time  Repetitions for statistical accuracy  Valid only for the parameters set used  Scalability of Simulation Area  Edge Effects  Shortcomings of Simulations  Substantial Processing Power and Time  Repetitions for statistical accuracy  Valid only for the parameters set used  Scalability of Simulation Area  Edge Effects

9. Dropped Packets  Probability of Dropping a Packet  Capacity per Cluster:  Number of Data Slot per Frame  Nonlinear relation between Load and P dp  Detailed probability distribution of Load is needed

10. Dropped Packets  P s : Ratio of number of nodes in spurt to all nodes  Voice Activity Detector  N CH : Number of CHs each node can receive access from  N CM : Number of nodes in the Cluster

11. Dropped Packets  Effect of Dropped Packets on Throughput  Considering Rx Throughput  Each node  all one hop neighbors

12. Collisions  Number of frames (N f ) vs. co-frame CH separation (d ch )  Labeling structure used in cellular systems  Co-frame CH separation (d ch ) vs. number of collisions ( f coll )  Correlation between  Number of Nodes that can cause collisions  Number of Collisions

13. Agenda  Soft Clustering Approaches  Protocol Overview  Analytical Model  Dropped Packets  Collisions  Proof of Concept  Optimization of TDMA parameters

14. Proof of Concept  Number of Packets Lost per Superframe (N f = 6)

15. Proof of Concept  Number of Packets Lost per Superframe (N f = 8)

16. Agenda  Protocol Overview  Analytical Model  Dropped Packets  Collisions  Proof of Concept  Optimization of TDMA parameters

17. Optimization of TDMA parameters  optimization with corresponding throughput figures with respect to the maximum realizable throughput Theoretically Optimized N f

18. Conclusions and Future Work  The model  Accurate  Can be used in optimization of parameters  Instantaneous results for changing  Transmission Power  Propagation Model  PHY Specs: Freq, Threshold values …  Asymptotic behavior  Energy consumption  Average node sleep/awake durations  Average energy consumption per node  Node and CH comparison wrt energy consumption  Optimization of Nf wrt energy consumption  We are going to add effects of upper layers into the model

19. Thanks! Questions&Comments? Contact Info: Web: www.ece.rochester.edu/~karaoglu/www.ece.rochester.edu/~karaoglu/ E-mail : karaoglu@ece.rochester.edukaraoglu@ece.rochester.edu Thanks! Questions&Comments? Contact Info: Web: www.ece.rochester.edu/~karaoglu/www.ece.rochester.edu/~karaoglu/ E-mail : karaoglu@ece.rochester.edukaraoglu@ece.rochester.edu

20. PHY Layer Abstraction  BW  Each tx occupies some part of the BW  Transmissions should overcome any noise present in the space of the BW  Divide  Spatial reuse

21. PHY Layer Abstraction  TDMA:  Divide BW along time axis  Clustering:  Distribute parts of BW spatially among clusters

22. Protocol Overview  TDMA  Soft Clustering  CHs responsible for channel access only  Inter cluster communication is allowed

23. Analytical Analysis  Shortcomings of Simulations  Substantial Processing Power and Time  Repetitions for statistical accuracy  Valid only for the parameters set used  Scalability of Simulation Area  Edge Effects

24. Analytical Analysis  Factors limiting performance:  Dropped Packets  Real-time communication  Limited Local Capacity  Clustering  Uneven distribution of Load  Node Distributions  Mobility  Collisions  Spatial Reuse  Limited BW Divisions

25. Dropped Packets  Probability of Dropping a Packet  Capacity per Cluster:  Number of Data Slot per Frame  Nonlinear relation between Load and P dp  Detailed probability distribution of Load is needed

26. Dropped Packets  p s : Probability of a node to be in spurt duration  p A : Probability of a node to be in the communication range of a CH  p d : Probability of a node that is in the communication range of a CH to choose that CH as its channel access provider  Independent of Node Density  assumed constant

27. Dropped Packets  p s : Probability of a node to be in spurt duration  p A : Probability of a node to be in the communication range of a CH  p d : Probability of a node that is in the communication range of a CH to choose that CH as its channel access provider p dn = p s p A p d

28. Collisions  Number of frames (N f ) vs. co-frame CH separation(d ch )  Labeling structure used in cellular systems

29. Collisions  co-frame CH separation (d ch ) vs. number of collisions ( f coll )

30. Collisions  co-frame CH separation (d ch ) vs. number of collisions ( f coll )  N nCH : Expected number of nodes in the cluster  N n : Total number of nodes  N C : Number of cluster in 2*r comm range  V : Region bounded by the circle with radius 2*r comm around origin  fcoll : number of packets lost per SF due to collision

31. Agenda  Soft Clustering Approaches  Protocol Overview  Analytical Analysis  Dropped Packets  Collisions  Proof of Concept  Optimization of TDMA parameters

32. Proof of Concept  Total Number of Packets Lost per Superframe

33. Proof of Concept  RX Throughput per Superframe

34. Agenda  Soft Clustering Approaches  Protocol Overview  Analytical Analysis  Dropped Packets  Collisions  Proof of Concept  Optimization of TDMA parameters

35. Optimization of TDMA parameters

36. Other Uses of the Model  Instantaneous Analysis Results for changing  Transmission Power  Propagation Model  PHY Specs: Freq, Threshold values …  Asymptotic Behavior  Energy Consumption  Average node sleep/awake durations  Average energy consumption per node  Node and CH comparison wrt energy consumption  Optimization of Nf wrt energy consumption

37. Throughput Per Node

38. Energy Consumption per Node

39. Thanks! Questions&Comments?