1. Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing Cristina Rico García www.DLR.de Chart 1Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12 Promotionsvortrag Ulm, 18/10/2012 TexPoint fonts used in EMF. Read the TexPoint manual before you delete this box.: A AAA A A A A A A 

2. Motivation 1.IntroductionM New generation of safety systems: Mobile Ad-hoc Networks Using Beaconing. (Beaconing MANETs) Challenge: Coordination of transmissions. Goal: Robust Medium Access Control (MAC) protocol. www.DLR.de Chart 2>Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12 50000 traffic fatalities in 2010 in EU-27 Robust and efficient MAC protocol

3. Outline 1.Update Delay: Novel metric for MAC protocols in beaconing MANETs 2.COMB: New two-layered protocol 3.Case Study: MAC design for the Railway Collision Avoidance System www.DLR.de Chart 3> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12

4. Outline 1.Update Delay: Novel metric for MAC protocols in beaconing MANETs 2.COMB: New two-layered protocol 3.Case Study: MAC design for the Railway Collision Avoidance System www.DLR.de Chart 4> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12

5. Case Study: Need for the Update Delay www.DLR.de Chart 5> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12 Entropy PDF Position time ??? The Update Delay has a direct relation to the application performance. UD=4 frames

6. Ideal CCDF(UD k ) Probability that it takes more than 6 frames to receive a beacon is 10 -5 A New Metric: The Update Delay www.DLR.de Chart 6> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12 Probability that the update delay is longer than k frames # frames ( k ) beaconing frame duration CCDF(UD k ) Required performance with lower capacity

7. Performance Analysis with Update Delay Performance Factors Observed offered traffic G : [% of capacity] Uncertainty MAC information H(S MAC ) : entropy Hidden nodes Beaconing frame duration T f : [s] Renewal cycle r : [# frames] Capture : [dB] Interference I : [% of capacity] Protocol parameters p These factors are sufficient to describe the main effects of MAC, incl. capacity, # nodes, communication range, path loss, distance, speed, etc. www.DLR.de Chart 7> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12

8. Self Organized TDMA (SOTDMA) Influence of Hidden Terminals on the Update Delay: Slotted. Commercially used in AIS. Reservation of next slot. www.DLR.de Chart 8> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12 16-fold capacity for SL=10 -4 due to hidden nodes # frames ( k ) CCDF(UD k ) No Interference

9. Optimisation of MAC Protocol Parameters Difference between Throughput and Update Delay www.DLR.de Chart 9> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12 Minimum CCDF(UD) Maximum Throughput (Saturated Throughput) Observed Offered Traffic Design optimisation for: Beaconing rate, Capacity, Communication Range CCDF(UD)| G f Throughput Slotted Aloha protocol 0.5 0 0.367

10. Carrier Sensing Multiple Access (CSMA) Optimisation by Means of Update Delay or Throughput Random access. Standard for vehicular networks communication. Backoff. www.DLR.de Chart 10> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12 10-fold gain in capacity for SL=10 -4 Update Delay (s) 1-Dimensional Interference Network Optimisation by means of Update Delay Optimisation by means of Throughput

11. Outline 1.Update Delay: Novel metric for MAC protocols in beaconing MANETs 2.COMB: New two-layered protocol developed in this thesis. 3.Case Study: MAC design for the Railway Collision Avoidance System www.DLR.de Chart 11> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12

12. Two-level protocol: 1.A channel is assigned to every cell depending on its position. Hidden nodes transmitt in different channels: COMB condition: D cell <R c. 2.Inside cells the transmission moment is chosen dynamically with any protocol using MAC information (SOTDMA, CSMA). With SOTDMA the nodes reserve future slot and channel depending on the orientation. COMB: Cell-based Orientation-aware MAC Broadcast Eliminating Hidden Nodes and Interference www.DLR.de Chart 12> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12 RcRc D cell B BA C B C A time

13. COMB for Two-dimensional Interference Networks www.DLR.de Chart 13> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12 1 2 3 4 5 10 11 1 5 6 7 8 9 10 11 12 3 4 5 6 7 11 12 1 Elimination of hidden nodes: Elimination of first interference zone: COMB condition: Number of cells to reuse distance relation:

14. Comparative Evaluation of Results www.DLR.de Chart 14> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12 2-Dimensional Interference Network 7-fold gain in capacity for SL=10 -4 Normalised Update Delay to G f =1 CCDF(UD k )

15. COMB for One-dimensional Interference Networks 1 www.DLR.de Chart 15> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12 5671432345 COMB condition: Number of cells to reuse distance relation: Elimination of hidden nodes: Elimination of interference zones: 1 276

16. Outline 1.Update Delay: Novel metric for MAC protocols in beaconing MANETs 2.COMB: New two-layered protocol 3.Case Study: MAC design for the Railway Collision Avoidance System www.DLR.de Chart 16> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12

17. Case Study: a Suitable Protocol for RCAS www.DLR.de Chart 17> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12 Worst-case scenario: Highest density in larger area  Shunting Yard „Maschen“ Statistics of accidents: 2138 accidents in EU-27 in 2010  40 kbits/s T f = 11.25 s 122 kbits/s T f = 4.5 s 200 kbits/s T f = 3.75 s CCDF(UD k ) Normalised Update Delay to G f =1 1-dimensional Interference Network

18. Conclusions 1.The Update Delay metric should be used instead of the traditional throughput. 2.The protocols must be designed to work with the optimum observed offered traffic. 3.The COMB protocol should be used instead of CSMA. It needs a tenth of capacity for the same performance. www.DLR.de Chart 18> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12

19. Contributions and Future Work Contributions 1.Novel Update Delay metric. 2.Analysis methodology. 3.Optimisation methodology. 4.Study of Slotted Aloha, SOTDMA, CSMA. 5.Novel protocol COMB. 6.Design of MAC for RCAS. Future Work 1.Results with detailed channel model. 2.Results with different interference zones. 3.Adaptive MAC protocols. www.DLR.de Chart 19> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12

20. Backup www.DLR.de Chart 20> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12

21. Conclusions 1.Update Delay: a.Comprises collision, reception rate, channel access delay. b.Effect of random and periodical transmissions. c.Highest impact have interference and hidden nodes. Other factors negligible. d.G Optimum ¼ 0.2 for all protocols. 2.Slotted protocols show superior performance with interference. 3.Optimum design with update delay can improve efficiency by the 10-fold. 4.Optimum COMB improves efficiency with respect to optimum CSMA by the 10-fold and by the 100-fold with respect to non-optimised CSMA. www.DLR.de Chart 21> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12

22. Backup Outline 1.. www.DLR.de Chart 22> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12

23. Slotted Aloha Analysis by Means of Update Delay Historical slotted protocol. Benchmark. Not influenced by hidden nodes. www.DLR.de Chart 23> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12 # frames ( k ) CCDF(UD k ) No Interference

24. Comparative Evaluation of Results www.DLR.de Chart 24> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12 Closed Network Hidden Nodes 1-dimensional Interference Network 2-dimensional Interference Network Gain in capacity

25. Channel Model

26. Path Loss Doppler Fading Delay Spread Frequency dependence Characterization of a Channel Model m dB Path loss: 10n log10(d) +C; n=2-4 Fast fading. Doppler Slow fading Analysis for each Scenario

27. Frequency Selection Frequency Bands for Railways in the world: USA and Canada: 159.810 – 161.610 and 452.900-457.9625 MHz Japan: 300 MHz band. Europe: GSM-R  876 - 880 MHz and 921 - 925 MHz Likely to be allocated to international railways  456 - 459 MHz and 460 - 470 MHz Restricted Bandwidth (12 KHz) 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 0 90 0 MHz USA and Canada Selected: 400 MHz Frequency Band We share the band with other systems  Protect the signal Japan European international railways GSM-R

28. Scenarios Train Station Scenario Area smaller than 1 km 2. Formed by parallel „streets“ separated by platforms with or without roofs that may interfere the communication but that are situated either well above the antenna level or under it. Therefore, train stations allow line of sight. They are similar to wide avenues or parallel streets in urban areas The train speed is typically under 20 km/h. Shunting Yard Scenario Open areas, smaller than 1 km 2. Grouping parallel rails and usually close to train stations and delimited by cuttings. Most of the structures in the shunting yards are metallic and under the train antenna level. Thus, line of sight is possible. The maximum speed is typically between 25 km/h and 40 km/h. Regional Network Scenario Covers areas larger than 1 km 2. Narrow clean area ≥ 11,6 m with curve radius down to 160 m that do not always allow line of sight. It might be surrounded by structures in both sides like tunnels, cuttings, and forest which may have a guiding effect. Bridges are not a source of shadowing since they are above the antenna level. Mountains may cause severe shadowing. Speeds up to 200 km/h are possible.

29. Channel Models for the scenarios Train Stations Path Loss: Kaji and Akeyama microcell model in wide avenues with low antenna levels applicable. Doppler: Speed < 20 km/h  Maximum Doppler shift: Few Hz Line of Sight  Ricean Doppler spectrum (COST 207). Fading: Line of Sight  Rice fast fading channel. K = - 1.2 dB Delay Spread: 1.6 – 5 μs. Shunting Yards Path Loss: Two ray model for microcells is applicable. Doppler: Speed = 20 km/h - 40 km/h  Maximum Doppler shift: Few Hz Line of Sight  Ricean Doppler spectrum. (COST 207). Fading: Line of Sight  Rice fast fading channel. K = - 1.2 dB Delay Spread: 9 μs Regional Networks Path Loss: Rural Hata-Okumura Model (Suburban in presence of mountains). Doppler: Speed < 200 km/h  Maximum Doppler shift: 148 Hz No Line of Sight is guaranteed  Jakes Doppler spectrum. Fading: No Line of Sight is guaranteed  Rayleigh fast fading. σ = 6 dB Delay Spread: Open area: 0.4 - 0.6 μs Mountainous area: 20 μs. In tunnels the attenuation is 15-20 db/km plus 15-20 db at the entrance and exit. Shadowing caused by narrow structures like bridges is negligible, in the order of 2-5 dB.

30. RCAS

31. Advantages and drawbacks of railway transport Railway infrastructure market volume in Europe Road transport Energy consumption Cost Railway transport Safety and traffic management

32. Traditional safety and traffic management in railway Signals passed at a danger (SPAs): one per day in UK, 21% dangerous situations. E.g. accident Oakland, October 2011, 18 injured. Several hardware components: higher failure probability and robbery. E.g. accident Wenzhou, July 2011, 40 killed. Cost of caused delay in UK: US $61 million for 12600 hours. Unnefficient: Time and energy los because of unnecessary stops at signals.

33. Nature and transport Nature can inspire a radically different approach: Self organizing Members adapt „on the fly“ to changing environment Cooperation among members Simple action of individual members Complex aggregate behaviour Behaviour based on organisms‘ intrinsic „pre-programmed“ parameters + environmental situation. Self organizing approach inspired by nature to control railway traffic: Two level algorithms

34. First level algorithm Set timer according to beaconing rate Send status beacon at MAC layer Is the timer at zero? Coordinate speed with sender No Yes Receive message? No

35. Second level algorithm Train monitors distance to conflict zone while approaching Train detects entrance to conflict zone Train broadcasts message asking for network status Receives answer from leader? Coordinates with leader Train is the new leader Receives message asking for status? Answers with status information Train is not the leader anymore Receives message asking/answer for status, conflict zone larger than communication range and own position closer to opposite end of conflict zone than the message creator? Rebroadcast message with the slotted 1 persistence broadcast scheme Train still inside the conflict zone? No Yes No Yes No

36. Implications. Hardware cost savings 1. Radio with antennas (eg. Balises) : US $700 each 2. Fiber to connect the radios: US $14000 per kilomenter 3. Regional computers: US $100000 each US $ 30000 per kilometer only in hardware

37. Implications. Energy savings Fraction of energy not recouped by regenerative brakes Mass of the train Speed before braking Adapted speed to next train

38. Implications. Time savings Friction coefficient Locomotive acceleration force Earth acceleration... T 1 Braking coefficient Mass of the train Speed before braking Adapted speed to next train

39. Implications. Total cost savings 1. Infrastructure savings: US $ 30000 per kilometer 2. Energy savings: US $ 22 per unnecessary stop, for average train and energy cost US 15 cents per kWh. 3. Time savings: US $ 327 per unnecessary stop, for 200 passengers US $ 40 wage per hour and US $ 80 for the railway company. Example: UK730000 train stops per year 21% of the stops neccessary to avoid danger 32000 km rail tracks Maximum: 400 TJ and 18655 hours lost  US $ 254.77 millions per year. Minimum: 84 TJ and 3917 hours  US $ 53.3 millions per year. Savings of modernizing safety and traffic management: US $ 960 millions

40. Conclusions Reducing cost of safety and traffic management. Increasing the energy and time efficiency of railway transportation by mitigating the number of accelerations and decelerations. Supporting a greener environment by eliminating most infrastructure-based safety and traffic control elements. Enabling ubiquitous traffic control and increasing safety in „dark territories“

41. Case Study: a Suitable Protocol for RCAS www.DLR.de Chart 41> Adatptive and Tractable Bayesian Context Inference for Resource Constrained Devices> Korbinian Frank > 25/07/12 Worst case scenario: Highest density in larger area.  Shunting Yards. Approximated by 1-Dimenstional interference Network. Bailey Yard, USA: Length: > 13 km Width: 3.2 km Dailiyrolling stocks: 10000 Maschen, Germany: Length: > 6 km Width: 0.7 km Dailiyrolling stocks: 11000

42. Case Study: a Suitable Protocol for RCAS www.DLR.de Chart 42> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12 Communication range: 4 ¢ d brake ¼ 5 km Worst-case scenario: Highest density in larger area  Shunting Yards Maschen, Germany: Length: > 6 km Width: 0.7 km Density: 100 Statistics of accidents: 2138 accidents and 7113 Signals passed at a danger (SPAs) in EU-27 in 2012  40 kbits/s 122 kbits/s 200 kbits/s

43. www.DLR.de Chart 43> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12

44. State of the Art

45. Protocols Clasification

46. Information Theory approach

47. Prediction and Update Steps

49. Entropy in Prediction and Update Steps

50. Metrics

51. Packet delivery rate and throughput

53. A New Network, A New Metric: The Update Delay www.DLR.de Chart 53> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12 beaconing frame duration # frames # nodes in the network reception probability at node j from node i at frame k collision probability at frame k Probability that the update delay is longer than k frames

54. Performance Factors

55. Performance Analysis www.DLR.de Chart 55> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12 Observed offered traffic Uncertainty MAC information Hidden nodes Beaconing frame duration Renewal cycle Capture Interference Protocol parameters Capacity Beacon length # nodes in communication range speed Communication range Node2node distance Path loss exponent # nodes out of communication range Hidden nodes Beaconing frame duration Renewal cycle Interference

56. Performance Analysis with Update Delay www.DLR.de Chart 56> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12 Performance factors: v p(UD)=f(v1,…, v8) System, application, and environmental parameters Bits, capacity, # nodes Renewal cycle, beaconing Speed, comm. range Nodes distance, path loss Nodes in network Observed offered traffic Uncertainty MAC information Hidden nodes Beaconing frame duration Renewal cycle Capture Interference Protocol parameters

57. Importance of the Transmission Moment Performance Factors: Hidden Nodes, Renewal Cycle and Beaconing Frame Duration www.DLR.de Chart 57> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 18/10/12 # frames ( k ) CCDF(UD k ) meters Renewal cycle=0 Constant beaconing rate Renewal cycle>0 Constant beaconing rate Renewal cycle>0 Constant beaconing rate

58. Interference www.DLR.de Chart 58> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12

59. Results on State of the Art Protocols Five scenarios: 1.Closed Network 2.Hidden Terminals 3.Capture 4.1-Dimension Interference 5.2-Dimension Interference 6.Factors: G, r, T f, p, TM, ® www.DLR.de Chart 59> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12 Three State of the Art MAC Protocols: 1.Slotted Aloha: Slotted, historial, benchmark. No affected by hidden nodes. 2.Self Organized TDMA: Commercially used in AIS. Ideal performance in closed networks. 3.Carrier Sensing MA: In the IEEE 802.11p standad for vehicular networks. Congests for low G. m dB m m m m

60. Optimization

61. Optimum Communication range www.DLR.de Chart 61> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12

62. Robust and Efficient MAC Protocols 1.Higher performance for a given capacity: 2.Lower capacity for a required performance: www.DLR.de Chart 62> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12 Maximum Throughput Minimum CCDF(UD) Best performing protocol Key: Fair comparison Design: Beaconing Rate Capacity Comm. Range: 4 ¢ d brake

63. Protocol Optimisation Worst-case application requirement: UD max =4 s at CCDF(UD)=10 -5 www.DLR.de Chart 63> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12 Full Network Optimum curve

64. Slotted Aloha

65. Slotted Aloha Flowchart www.DLR.de Chart 65> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12

66. Slotted Aloha Protocol Clasification www.DLR.de Chart 66> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12

67. Slotted Aloha Collision and Reception Probabilities www.DLR.de Chart 67> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12

68. A New Network, A New Metric: The Update Delay www.DLR.de Chart 68> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12 beaconing frame duration # frames # nodes in the network reception probability at node j from node i at frame k collision probability at frame k Probability that the update delay is longer than k frames

69. Slotted Aloha Update Delay www.DLR.de Chart 69> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12

70. SOTDMA

71. SOTDMA Flowchart www.DLR.de Chart 71> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12

72. SOTDMA Protocol Clasification www.DLR.de Chart 72> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12

73. CSMA

75. CSMA protocol classification

76. Throughput, one dimensional interference

77. COMB

78. COMB: Cell-based Orientation-aware MAC Broadcast Eliminating Hidden Nodes and Interference www.DLR.de Chart 78> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12 Two-level protocol: 1.Upper level: hidden nodes transmitt in different upper channels. Condition: D cell <R c. 2.Lower level: Inside the upper channels state of the art protocols using MAC information (SOTDMA, CSMA) RcRc D cell

79. COMB for Two-dimensional Interference Networks www.DLR.de Chart 79> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12 1 2 3 4 5 10 11 1 5 6 7 8 9 10 11 12 3 4 5 6 7 11 12 1 Elimination of hidden nodes: Elimination of first interference zone:

80. COMB for One-dimensional Interference Networks 1 www.DLR.de Chart 80> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12 21211212121 Eliminate hidden nodes: Eliminate first interference zone:

81. COMB for Two-dimensional Interference Networks www.DLR.de Chart 81> Adatptive and Tractable Bayesian Context Inference for Resource Constrained Devices> Korbinian Frank > 25/07/12 1 2 3 4 5 10 11 1 5 6 7 8 9 10 11 12 3 4 5 6 7 11 12 1 Eliminate hidden nodes: Eliminate first interference zone:

82. COMB for One-dimensional Interference Networks 1 www.DLR.de Chart 82> Adatptive and Tractable Bayesian Context Inference for Resource Constrained Devices> Korbinian Frank > 25/07/12 21211212121

83. Channel division www.DLR.de Chart 83> Adatptive and Tractable Bayesian Context Inference for Resource Constrained Devices> Korbinian Frank > 25/07/12 1-dimensional Interference Network 2-dimensional Interference Network

84. Statistics

85. Statistics of type of accidents www.DLR.de Chart 85> Adatptive and Tractable Bayesian Context Inference for Resource Constrained Devices> Korbinian Frank > 25/07/12

86. Statistics of accidents precursos www.DLR.de Chart 86> Adatptive and Tractable Bayesian Context Inference for Resource Constrained Devices> Korbinian Frank > 25/07/12

87. Synchronization

88. Length of Messages www.DLR.de Chart 88> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12 Message length: Safety information needs around 200 bits. Header length: Slot reservation: Number of bits for: Number of slots per frame * maximum number of frames until next transmission. Channel reservation: 3 bits for the one dimensional interference network and 4 bits for the two-dimensional interference network.

89. Sincronización www.DLR.de Chart 89> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12 Length of a slot: B=40 kbits/s Bits per beacon= 200 bits  Bits per beacon/B= 5 ms Frame length=11.25 s Nodes density=100 trains/km2 Nodes per cell in Maschen=100*0.7*3.6=250 Numer of nodes per cell= GPS accuracy in computer +/-0.1 s  <<1% of the beacon length in time

90. Most efficient MAC protocol for a given capacity www.DLR.de Chart 90> Adatptive and Tractable Bayesian Context Inference for Resource Constrained Devices> Korbinian Frank > 25/07/12 MAC protocols Performance Predictor MAC Comparator MAC 1 MAC 2 MAC 3 MAC 4 MAC N CCDF(UD) MAC1 CCDF(UD) MAC2 CCDF(UD) MAC3 CCDF(UD) MACN v1v1 v2v2 v3v3 vLvL csp 1 csp 2 csp 3 csp M Z Performance Factors Calculator MAC Optimum CCDF(UD) MACOptimum

91. Most efficient protocol and system parameters for a given application www.DLR.de Chart 91> Adatptive and Tractable Bayesian Context Inference for Resource Constrained Devices> Korbinian Frank > 25/07/12 MAC protocols System Parameters Optimizer Bandwidth Comparat or MAC 1 MAC 2 MAC 3 MAC 4 MAC N cp 1, cp 2 … cp L Z Application Requirements Calculator App sp I+1 sp I+2 sp M CCDF(UD max ) UD max sp 1, sp 2 … sp I (sp I+1, sp I+2 … sp M ) MACN (sp I+1, sp I+2 … sp M ) MAC1 (sp I+1, sp I+2 … sp M ) MAC2 MAC Optimum

92. Outline 1.Introduction 2.Bayeslets: Modular Bayesian Inference Rules 3.A Bayeslet for the Estimation of Symbolic Location 4.A Bayeslet for Human Motion Related Activity Recognition 5.Contributions www.DLR.de Chart 92> Adatptive and Tractable Bayesian Context Inference for Resource Constrained Devices> Korbinian Frank > 25/07/12

93. Beaconing MANETs vs. Traditional MANETs Tx Rx Network Classical network t=transmission delay buffer Network Video, sound, data… Tx Beaconing MAMNETs Short status update May I send? MAC Rx t=transmission delay~0 No Yes!

94. Problem Statement www.DLR.de Chart 94> Medium Access Control Protocols in Mobile Ad-hoc Networks Using Beaconing> Cristina Rico > 26/09/12 Entropy PDF Position 1 time max Entropy ???