1. SOS: A Safe, Ordered, and Speedy Emergency Navigation Algorithm in Wireless Sensor Networks Andong Zhan ∗ †, Fan Wu ∗, Guihai Chen ∗ ∗ Shanghai Key Laboratory of Scalable Computing and Systems, Department of Computer Science and Engineering, Shanghai Jiao Tong University, China †Department of Computer Science, Johns Hopkins University, USA IEEE ICCCN (Accepted rate:29.6%)

2. Wireless & Mobile Network Laboratory Outline  Introduction  Related Works  Problem Specification  Design Principles  Simulations  Conclusions 2

3. Wireless & Mobile Network Laboratory Introduction  More people were killed by natural disasters worldwide past.  Evacuation techniques are highly needed to navigate the personnel out of danger quickly.  Wireless sensor networks can play an important role in detecting disasters and navigating the personnel out of dangerous areas. 3

4. Wireless & Mobile Network Laboratory Introduction  Objective The objective of a successful navigation is to schedule all the users to bypass the dangerous areas safely, and finally evacuate to pre-known exits as soon as possible.  Requirements All the escape paths given by the navigation algorithm should be safe paths. All the users should evacuate from the emergency area orderly without congestion. The navigation algorithm should minimize the total evacuation time. To lower the cost, sensor nodes do not require geographical location information. 4

5. Wireless & Mobile Network Laboratory Problem Specification 5

6. Wireless & Mobile Network Laboratory Problem Specification  Assumptions a wireless sensor network can detect dangerous areas Users carry wireless communication devices which enable them to “talk” with nearby sensors e.g., 802.15.4 compatible PDA or smart phone 6

7. Wireless & Mobile Network Laboratory Problem Specification  Safe Path: A path P = {s 1, s 2,..., s n } is a safe path if and only if ∀ s i ∈ P, d i ≥ d Γ d i is the distance between sensor node s i and its nearest alarming neighbor node. d Γ is the safe distance threshold.  Safety Capacity: For each sensor node s i on a safe path, its safety capacity is the maximum number of people passing through it safely per unit time. 7

8. Wireless & Mobile Network Laboratory Design Principles  Constructing the Medial Axis Graph  Formulating the Navigation Schedule Problem  Designing the Distributed Algorithm 8

9. Wireless & Mobile Network Laboratory Design Principles  Constructing the Medial Axis Graph  Formulating the Navigation Schedule Problem  Designing the Distributed Algorithm 9

10. Wireless & Mobile Network Laboratory Design Principles 10 Constructing the Medial Axis Graph

11. Wireless & Mobile Network Laboratory Design Principles  Constructing the Medial Axis Graph  Formulating the Navigation Schedule Problem  Designing the Distributed Algorithm 11

12. Wireless & Mobile Network Laboratory Design Principles  Navigating users to the closest gateway, choosing the safest path for every user, or taking both safety and distance into account  When the number of users is large and the capacities of safe paths are low, congestion may occur, and greatly increases the evacuation time resulting in more casualties. Time should be considered. Waiting is inevitable when the number of users is larger than the maximum safety capacity of the network. 12 Formulating the Navigation Schedule Problem It can not guarantee the optimal scheduling

13. Wireless & Mobile Network Laboratory Design Principles 13 Formulating the Navigation Schedule Problem Time cost (between two nodes) = 1 Time cost=2 Time cost=3 Safety Capacity

14. Wireless & Mobile Network Laboratory Design Principles 14 Formulating the Navigation Schedule Problem Time cost (between two nodes) = 1 Time cost=2 The number of users is larger than the safety capacity Time cost=3

15. Wireless & Mobile Network Laboratory Design Principles  We create a graph G(V, E) and denote vertex i t ∈ V, i ∈ N as the state of sensor node s i at time t, and directed edge edge (i, j, t) ∈ E as the arc from vertex i t to vertex j t+1  A sensor node may have several vertices so as to represent the number of users in different time units. 15 Formulating the Navigation Schedule Problem it it edge (i, j, t) j t+1 it it edge (i, i, t) i t+1 Traveling:Waiting:

16. Wireless & Mobile Network Laboratory Design Principles 16 Formulating the Navigation Schedule Problem

17. Wireless & Mobile Network Laboratory Design Principles  The navigation schedule problem can be formulated as a linear program: 17 Formulating the Navigation Schedule Problem Minimize Subject to: the flow from vertex i t to j t +1 time cost the number of users in the whole safe area safety capacity

18. Wireless & Mobile Network Laboratory Design Principles  Constructing the Medial Axis Graph  Formulating the Navigation Schedule Problem  Designing the Distributed Algorithm 18

19. Wireless & Mobile Network Laboratory Design Principles 19 Designing the Distributed Algorithm

20. Wireless & Mobile Network Laboratory Design Principles 20 Designing the Distributed Algorithm i hopd=10 u(i) i hopd=3 u(i)

21. Wireless & Mobile Network Laboratory Design Principles 21 Designing the Distributed Algorithm S4S4 S3 S3 di:di: S2S2 S1S1 gateway Exit inoutinoutinout inout height: 4-hop3-hop2-hop1-hop 86427531 u(i) u(4)=10u(3)=5u(2)=10u(1)=15

22. Wireless & Mobile Network Laboratory Design Principles  Local Minimum Problem 22 Designing the Distributed Algorithm S4S4 S3 S3 S2S2 S1S1 height 8 6 4 2 4+1=5 Record the navigation schedule, i.e, the time and the number of users to certain neighbor node

23. Wireless & Mobile Network Laboratory Simulations  Simulate randomly deploying sensor nodes in a square field.  The number of nodes is from 1000 to 8000.  We randomly create 1 to 5 groups of users in each run.  The number of users in a group is created randomly between 1 and 50.  In each run, we also randomly setup 1 to 5 exits and 3 to 6 dangerous areas. 23

24. Wireless & Mobile Network Laboratory Simulations  Uniform capacity: the capacity of every sensor node is equal.  Linear capacity: the capacity of a node is linear with the number of hops from the node to the closest alarming node. 24 Linear capacity:

25. Wireless & Mobile Network Laboratory Simulations  Average evacuation time 25 Potential Feld (PF) Skeleton Graph (SG) Road Map (RM)

26. Wireless & Mobile Network Laboratory Simulations  Last evacuation time 26

27. Wireless & Mobile Network Laboratory Simulations  Network overhead 27

28. Wireless & Mobile Network Laboratory Conclusions  We have proposed SOS emergency navigation algorithm in WSNs.  To minimize users’ evacuation time, we have converted the emergency evacuation problem to a traditional network flow problem and used push-relabel algorithm to solve it.  In simulations, SOS is better than existing approaches in terms of average evacuation time, last evacuation time, and network overhead. 28

29. Thanks for your attention ！