CACA Linkbased Channel Allocation Exploiting Capture Effect for

CACA: Link-based Channel Allocation Exploiting Capture Effect for Channel Reuse in Wireless Sensor Networks Junghyun Jun 1, Solchan Yeon 2, Titir Kundu 3, Dharma P Agrawal 3, Jaehoon 4 1 Indian Institute (Paul)of. Jeong Technology Ropar, 2 Kookmin University, Republic of Korea, 3 Department of EECS, University of Cincinnati, Ohio State, USA, 4 Department of Interaction Science, Sungkyunkwan University, Republic of Korea 1

Capture-effect: an event of receiving (capturing) a packet from a collision A A’s packets captured B Both packets are lost B’s packets captured 2

Packet capture is not solely dependent upon relative RSS Clean Packet: Preamble Sync Packet capture if Interference is weaker: Preamble Sync Data Interfering packet Data packet Interfering Single Antenna Packet capture if Preamble Sync packet Data Interfering Interference is much weaker: Preamble Interfering packet Sync Data Preamble Sync Interfering packet Data : May capture a packet in case of Message in Message (Mi. M), otherwise packet is lost, Independent of RSS An original image is from Bernhard Firner, Multiple Receiver Strategies for Minimizing Packet Loss in Dense Sensor Networks, Mobi. Hoc'10 3

Wireless communication areas that directly exploited capture-effect • Fast Flooding: Flashing Flooding by Lu et al. , 2009 • Fast Time synchronization: Glossy by Ferrari et al. , 2011 • Throughput Gain: Source. Sync by Rahul et al. , 2010, Exor by Biswas, 2005 • Fast bulk data dissemination: Splash by Doddavenkatappa et al. , 2013 • Fast All-to-all communication: Chaos by Landsiedel et al. , 2011 • Collision Detection by Whitehouse et al. , 2005 Another area: Channel Allocation 4

Performance Improvement by exploiting channel diversity A High interference S B C A Interference Free S B T D Single Channel Network C T D Multi-Channel Network 5

Are Interference Free Links Always Better? • Yes, if all the channels’ qualities are similar • In reality, a link quality varies across different channels Ch PRR 11 . 33 13 . 19 15 . 63 17 . 46 19 . 81 21 . 81 23 . 41 25 . 28 good [CC 2420 RF Transceiver, Manjunath et al. RTSS’ 2011] 6

Random Channel Selection May Give Negligible Improvement or Worse Performance High interference A S PRR = 1 / PRR with Interfere = 0. 2 B C A S PRR = 1 / PRR with Interfere = 1 B 1 T 1 Interference Free 0. 2 1 C Bad Channel T 1 1 D D Single Channel Network 4 -Channel Network Equivalent Overall Performance 7

How about reusing some good channels? Reduce interference High interference A S 1/0. 2 B C S 1/0. 5 B 1 T 1 Interference from B A C Single Channel Network 1 T 1 1 D 1 Reuse Blue Channel D 3 -Channel Network with channel reuse 8

Goal: Which Links Should Reuse Blue Channel? All Same? Case (A) A S 1/? ? B C Case (B) T C 1 D S 1/? ? B 1 1 A Case (C) T C 1 D S 1/? ? B 1 1 A 1 1 T 1 D 9

What will happen to channel reuse when capture-effect is considered? A S 1 Ps(A|B)=0. 5 B C Ps(A|D)=0. 9 1 T 1 Single Channel Network S 1/0. 9 B C 1 1 T 1 1 D Interference from B A D 3 -Channel Network with channel reuse 10

Detecting Capture-Effect Method 1 A Method 2 ACK_Packet. A S Packet. B B Overhear ACK_Packet. A Beacon [timestamp Packet. B] Overhearing based Capture-effect Detection, Yun et al. 2007 Cons: Overhead for detection Pream Sync Detection Data Pream Sync Crc Pream Data Crc Capture Detection by continuous preamble search, Whitehouse et al. 2005 Cons: Frame modification 11

CACA: Capture-effect Aware Channel Allocation First assign different channels to a pair of links gaining little benefit from capture-effect ① Select top K quality channels (e. g. , by PRR), C denotes a set of K channels ② Construct Capture Graph L(G) (Vertex-to-Edge dual graph (called Line graph)) ③ Do While nodes in L(G) with no channel assigned ④ Select an edge (u, v) in L(G) with maximum weight w(u, v) (trade-off) ⑤ Assign least utilized channels from C to both node u and v or to u or v whichever does not have a channel. ⑥ Remove edge (u, v) from L(G) ⑦ End While 13

Capture Graph: Truncated Vertex-to-Edge Dual Graph 1. Forwarding Tree A 2. Line Graph CT S DT B C 3. Capture Graph T CS BT CT DS DT D forwarding edge Interference edge AS ① Construct vertex-to-edge dual graph ② Remove edges between interference nodes ③ Connect two nodes sharing interference BT BS AS Interfering pairs of forwarding edges 14

Capture Graph: Assigning edge weight – ETX Tradeoff between two channels and one channel Capture Graph, L(G) Graph, G A CT S DT B C T BT The less edge weight means the more benefit from capture. D AS 15

Graph Coloring: First Assign Least Utilized Channel to Edge with Maximum Weight CT DT AS Channel 1 [0] [1] Channel 2 [0] [1] Channel 3 [0] [1] BT Which color for AS? Red is optimal solution, but the finding is NP-hard. Our algorithm uses Round Robin for color selection. 16

Channel Allocation: Convert Colored Capture Graph back to Original Graph with Corresponding Channels A CT DT BT B C Channel 1 Channel 2 AS S T D Channel 3 17

Experiment Results from 140 -node Indriya Testbed (Telos. B) with Top 3 channels More Capture-effect Avr. Degree = 5 Capture-effect can not help! Avr. Degree = 11 Interference increase [Eavedropping: Zhou et al. , Infocom, 2006] 18

Impact of available channels #packets sent = 98 nodes*100 unicast packets = 9800 Welsh-Powell Algorithm # Channels Used 3 5 7 9 11: Interferencefree Total packets received by BS 890 2946 3245 3382 1293 Observation 1: When 5 channels were used by Eavesdropping 1447 total packets were received by BS and average PRR was 0. 76 Key Observation: 5 good channels with good capture effect are better than an interference-free network with 11 channels with some poor channels. 19

• Contact: – Junghyun Jun: peterjun@iitrpr. ac. in – Jaehoon (Paul) Jeong: pauljeong@skku. edu 20