Multiple Receiver Strategies for Minimizing Packet Loss in Dense Sensor Networks

Transcription

1 Multiple Receiver Strategies for Minimizing Packet Loss in Dense Sensor Networks Bernhard Firner Chenren Xu Yanyong Zhang Richard Howard Rutgers University, Winlab May 10, 2011 Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

2 Building An Internet of Things System Assumptions Sensor Networks and The Internet of Things The Internet of Things envisions a world where everyday items have radios items communicate information this information makes our devices smart Our world has a lot of things in it If everything has a radio, could we hear anything over the noise? Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

3 Building An Internet of Things System Assumptions What Will These Networks Look Like? A warehouse is an example of a very dense deployment. 100s of item palettes. 1000s of individual items. Transmitters must be small inexpensive long lifetime Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

4 Building An Internet of Things System Assumptions Realistic Network Assumptions One-hop many-to-one topology is most practical Periodic packets Data streams vital, not single packets Energy constrained Short packets, ID + sensor data Control packets costly As many packets per second as possible (throughput) Not 100% delivery success All sensors must be heard within a time window Splitting a network into groups of clusters and cluster heads is practical. Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

5 Wireless Communication Protocols Let s Rethink Our Communication Protocol CSMA Uses receiver, reduces lifetimes Works poorly with small packets TDMA Control packet overhead reduces lifetimes No-Receiver Protocols Long lifetimes High throughput with short packets Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

6 Using Multiple Receivers Strategies Further Improving Scalability We have Power-efficient hardware Power-efficient protocols We need more capacity Even 100% efficient channel usage might not be enough We can add more channels This seems natural - use more receivers, increase capacity Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

7 Using Multiple Receivers Strategies We Know That Multiple Channels Increase Packet Throughput Regardless of the protocol Requires multiple receivers simultaneously operating on different frequencies Cellular networks Y-MAC Multichannel L-MAC Are multiple channels the most effective use of multiple receivers? Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

8 Using Multiple Receivers Strategies Multiple Receivers Can Increase Throughput in Several Ways On multiple channels (spectral diversity) Reduces contention and collisions On a single channel Packet combining is error correcting Only helps with errors from noise Increasing the capture effect (spatial diversity) Packets are received despite collisions Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

9 Using Multiple Receivers The Capture Effect Capturing Packets in a Collision Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

10 Using Multiple Receivers The Capture Effect Capturing Packets in a Collision Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

11 Using Multiple Receivers The Capture Effect Capturing Packets in a Collision Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

12 Using Multiple Receivers The Capture Effect Capturing Packets in a Collision Spatial diversity increases capture gains! Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

13 Using Multiple Receivers The Capture Effect Capture is not Solely Dependent Upon Relative RSS Reception also depends upon the relative timing of packets. A clean packet is split into three segments: Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

14 Using Multiple Receivers The Capture Effect Capture is not Solely Dependent Upon Relative RSS Interference in the data segment causes bit errors Same as on a noisy channel Error correction techniques can be used Will be received correctly if interference is weak Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

15 Using Multiple Receivers The Capture Effect Capture is not Solely Dependent Upon Relative RSS Interference during the sync word can cause receivers to miss the packet Error correction techniques still work on the data segment Interference must be weaker so that the sync can be properly decoded Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

16 Using Multiple Receivers The Capture Effect Capture is not Solely Dependent Upon Relative RSS With the preamble corrupted the receiver will have trouble distinguishing one packet from the other Otherwise this is similar to when the sync and data have interference. Will only be decoded if interfering packet is much weaker Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

17 Using Multiple Receivers The Capture Effect Capture is not Solely Dependent Upon Relative RSS If the interfering packet began transmitting first different problems occur The interfering packet has the receiver s attention Our packet will be ignored regardless of its relative signal strength A technique known as Message in Message (MiM) can improve this slightly Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

18 Testing the Capture Effect Tests to Quantify Capture Effect Gains Hardware Two wire-synchronized transmitters Two receivers for MiM Radio MHz 10 byte packet lasting 300 microseconds One transmitter varies its power over time Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

19 Testing the Capture Effect Observed Capture Rates (1-3) Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

20 Testing the Capture Effect MiM Helps Some Collision Types (4-5) Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

21 Testing the Capture Effect Combined Capture Probability Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

22 Testing the Capture Effect Predicting the Capture Effect How Often Does Capture Occur in Real Situations? We have the capture probability given relative signal strength Need the probability of different relative signal strengths Assume that transmitters are uniformly distributed Assume attenuation is a constant 1/r α Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

23 Testing the Capture Effect Capture with Two Transmitters Predicting the Capture Effect P capture with two transmitters can be described in terms of Transmitter to receiver distances L 1 and L 2 The attenuation over the distance between L 1 and L 1, = L 1 L 2 The relative distance for capture to occur, c Conversion from db to relative distance for c is: c = 10 /10α P capture = P (L 1 L 2 c) Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

24 Testing the Capture Effect Predicting the Capture Effect Capture with Many Transmitters We are assuming a uniform distribution for transmitters a is the minimum distance from the receiver to the transmitters If the receiver is mixed in with the transmitters then a = 0 b is the maximum distance from the receiver to the transmitters Integrate over the uniform pdf in the interval [a, b] to find the probability of the distance ratio c. Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

25 Testing the Capture Effect Predicting the Capture Effect Capture with Many Transmitters Integrating we get: b 1 1 b a a b a ( 1 cb 2 = (b a) 2 2 a = c 2 cx dy dx ab ca2 2 + a2 if a = 0. (1) Now we can evaluate how this impacts system performance. ) Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

26 Reaping the Benefits Collisions How Often do Collisions Occur? Sensors are transmit only Packet duration is δ seconds Periodic transmission every τ seconds Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

27 Reaping the Benefits Collisions How Often do Collisions Occur? Sensors are transmit only Packet duration is δ seconds Periodic transmission every τ seconds P 2 way collision = 2δ τ (2) P (collision N transmitters) = 1 P (no collision) N 1 = 1 (1 2δ τ )N 1 (3) Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

28 Reaping the Benefits Spatial Vs. Spectral Diversity Comparing Spectral and Spatial Diversity N transmitters C channels M receivers per channel With a single channel we have P no capture = (1 P capture ) M If transmitters are evenly divided between channels then P collision loss = 1 (1 P no capture (2δ/τ)) N/C 1 Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

29 Reaping the Benefits Spatial Vs. Spectral Diversity Comparing Spectral and Spatial Diversity Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

30 Reaping the Benefits Experimental Validation Experimental Validation 7 meter square testing area 100 transmitters in a uniform random distribution 8 receivers, each with a redundant receiver µsecond packets per second per sensor Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

31 Reaping the Benefits Experiment and Theory Agree Experimental Validation Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

32 Reaping the Benefits Experimental Validation Spatial Diversity Wins, Even Without MiM Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

33 Conclusions Reaping the Benefits Experimental Validation Multiple receivers on a single channel can outperform a system with the same number of receivers on multiple channels Once a certain number of receivers are used Especially in high attenuation environments Especially when packets are short Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

34 References Reaping the Benefits Experimental Validation Firner, B.; Xu, C.; Howard, R; Zhang, Y. Multiple Receiver Strategies for Minimizing Packet Loss in Dense Sensor Networks. MobiHoc 10, pp , September 20-24, Firner, B.; Jadhav, P.; Zhang, Y.; Howard, R.; Trappe, W.; Fenson, E. Towards Continuous Asset Tracking: Low-Power Communication and Fail-Safe Presence Assurance. Sensor, Mesh and Ad Hoc Communications and Networks, SECON 09. 6th Annual IEEE Communications Society Conference on, vol., no., pp.1-9, June 2009 Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

35 Reaping the Benefits Message In Message Brief Introduction to Message in Message Picture two transmitters and a receiver. Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

36 Reaping the Benefits Message In Message Brief Introduction to Message in Message Transmitter A begins sending a packet and the receiver starts decoding it. Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

37 Reaping the Benefits Message In Message Brief Introduction to Message in Message Without MiM: 1 Transmitter B begins transmitting a packet 2 A s packet (currently being decoded) is corrupted 3 B s packet is missed - the receiver was busy with A s packet. Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

38 Reaping the Benefits Message In Message Brief Introduction to Message in Message With MiM: 1 Transmitter B begins transmitting a packet 2 A s packet is corrupted. Simultaneously, the receiver begins decoding B s packet. 3 B s packet is decoded normally. Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

39 Reaping the Benefits Message In Message We Can Emulate Message in Message MiM is only found in some Wi-Fi devices. We can emulate it in sensor networks with the CC1100 Wire two radios together Switch to the redundant receiver when the first begins receiving a packet The redundant receiver decodes previously missed packets Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29

40 Reaping the Benefits Spatial Vs. Spectral Diversity Comparing Spectral and Spatial Diversity Bernhard Firner (Winlab) Multiple Receiver Strategies May 10, / 29