Maximizing Network Lifetime of WirelessHART Networks under Graph Routing

Transcription

1 Maximizing Network Lifetime of WirelessHART Networks under Graph Routing Chengjie Wu, Dolvara Gunatilaka, Abusayeed Saifullah*, Mo Sha^, Paras Tiwari, Chenyang Lu, Yixin Chen Cyber-Physical Systems Lab, Washington University in St. Louis Missouri University of Science & Technology * Binghamton University ^

2 Wireless for Process Automa1on Emerson 5.9+ billion hours operating experience 26,200+ wireless field networks $ million by 2020 [Market and Market] Offshore Onshore Courtesy: Emerson Process Management Killer App of IoT! 2

3 Industrial Wireless Challenges Ø Reliability Sensor Ø Real-time Ø Control performance Actuator sensor data Ø Energy efficiency: need long battery life in harsh environments! control command Controller 3

4 WirelessHART Ø Industrial reliability q Multi-channel TDMA MAC q Over IEEE PHY q Redundant routes Ø Centralized network manager q collects topology information q generates routes and transmission schedule q disseminates to field devices q re-computes routes when topology changes Industrial wireless standard for process monitoring and control 4

5 Graph Rou1ng Ø Handle link and node failures through path diversity Ø Graph route of a flow q a primary path q a backup path for each node on the primary path s 5 w 7 1, 2 u 3, 4 v 5, 6 d x 4 5 y z primary path backup path Ø Transmissions per hop q Two transmissions on the primary link dedicated TDMA slots q One transmission on the backup link shared CSMA/CA slot 5

6 Energy Cost of Reliability Ø Graph routing improves reliability at cost of energy Ø Measurement: +57% reliability at 1.7 energy compared to single-path source routing [EWSN'15] Graph routing: 88% Source routing: 31% 6

7 Challenges Ø Maximize network lifetime under graph routing q Industry demands multi-year battery life q Efficient routing in response to wireless dynamics Ø Unique challenges for WirelessHART networks q Centralized multi-path graph routing q Transmissions in dedicated and shared slots 7

8 Contribu1ons Ø Problem: network lifetime maximization under graph routing q Network lifetime = time till first node runs out of battery q NP hard Ø Three approaches q Optimal integer programming q Linear relaxation of the integer programming q Efficient greedy heuristic Ø Implementation on a WirelessHART testbed 8

9 Analyzing Power Consump1on Ø Model based on WirelessHART standard Ø 1-2 transmissions on primary path Ø 3 rd transmission on back path q Small probability, but receiver must turn on and listen. Ø Load: power consumption / battery capacity s 5 w 7 1, 2 u 3, 4 v 5, 6 d x 4 5 y z primary path backup path 9

10 Integer Programming Ø Objective: max min node lifetime à min max load Ø Graph route as constraints q An incoming primary link à an outgoing primary link q An incoming primary link à an outgoing backup link q An incoming backup link à an outgoing backup link s 5 w 7 1, 2 u 3, 4 v 5, 6 d x 4 5 y z primary path backup path Ø Optimal solution Ø High computational cost à cannot scale to large networks 10

11 Linear Programming Relaxa1on 1. Relax binary decision variables to real numbers 2. Linear Programming à real number solutions 3. Round real numbers to integer solutions based on threshold 4. Incrementally find the largest threshold with valid routes Ø Implemented in GNU Linear Programming Kit (GLPK) Ø Near optimal solution with affordable computational cost. 11

12 Greedy Heuris1cs Ø Compute routes for flows in the rate monotonic order Ø For each flow: find the graph route with minimum load q Load per node = power consumption / battery capacity q Incrementally add nodes with the smallest load to primary path and update neighbors load q Then select backup path with minimum load d Ø Iterate until no further improvement b c Ø Polynomial complexity s f e 12

13 Evalua1on Ø Implemented on a WirelessHART testbed (69 TelosB motes) q WirelessHART stack (multi-channel TDMA + routing) q Network manager (scheduler + routing) Ø Simulations based on testbed topology WUSTL wireless sensor-actuator network testbed 13

14 Compare to Op1mal (Small Network) Lifetime normalized to optimal solution from Integer Programming 10 nodes, 20 links GH & LP within 80% of optimal SP: Shortest Path RRC [Han 2011] GH: Greedy Heuristic LP: Linear Programming 14

15 Network Life1me (Testbed Topology) LP and GH lead to longer network lifetime 15

16 Execu1on Time (Testbed Topology) GH needs less time than LP 16

17 Conclusion Ø Industrial wireless networks is a killer app for IoT q Driven by industrial standards such as WirelessHART q Deployments rolling out world wide Ø Graph routing enhances reliability at high energy cost à energy efficiency is critical! Ø Three approaches to maximize network lifetime q Integer Programming: optimal q Linear Programming Relaxation: faster q Greedy Heuristic: fastest solution for run-time adaptation Ø Implemented with WirelessHART on testbed 17

18 Reading Ø C. Wu, D. Gunatilaka, A. Saifullah, M. Sha, P.B. Tiwari, C. Lu and Y. Chen, Maximizing Network Lifetime of WirelessHART Networks under Graph Routing, IEEE International Conference on Internet-of-Things Design and Implementation (IoTDI'16), April