1. SR: A Cross-Layer Routing in Wireless Ad Hoc Sensor Networks Zhen Jiang Department of Computer Science West Chester University West Chester, PA 19335, USA 12/7/2015Hong Kong PolyU

2. Outline Introduction Problem Our Approach Conclusion 12/7/2015Hong Kong PolyU

3. Introduction Routing problems in WASN applications Improvement on the entire routing path ◦ Length, delay, and performance ◦ Security, etc Topology information model ◦ Where link connections change dynamically ◦ For each relay at intermediate nodes Main factors ◦ Reliability, scalability, and cost effectiveness 12/7/2015Hong Kong PolyU

4. Existing routing schemes Centralized connection (1) Singe point of failure (2) Hot spots (energy depletion, interference, performance bottle neck, etc) (3) Low reliability (impossible for multi-hop relay in real applications) (4) Low scalability Not suitable in a highly dense and dynamic environment 12/7/2015Hong Kong PolyU

5. Problems 12/7/2015Hong Kong PolyU

6. Idea Solution 12/7/2015Hong Kong PolyU

7. Challenges Unpredictable configuration ahead due to ◦ Interferences ◦ Node failure ◦ Node mobility ◦ Privacy and selfishness ◦ Signal strength and energy consumption ◦ Traffic jamming Huge cost in probing to catch the configuration change ◦ Delay ◦ Information storage ◦ Computational cost 12/7/2015Hong Kong PolyU

8. Observations Reactive information model ◦ Not suitable for routing in dynamics Passive information model ◦ Hard to find an effective description for various pair of the source and destination Information Scale ◦ The farther the relay node to the destination, the less accurate information is needed. ◦ 1 -hop direct connection + k-hop reachability information 12/7/2015Hong Kong PolyU

9. Problem A new information model ◦ Indicate the neighbor preference for a 1-hop decision with the global path optimization  Existence of such a preference? ◦ Constructed in a passive information model,  How to keep relatively stable after dynamic changes (reliability when link changes and positions of source and destination change)? ◦ Minimize the construction process within a limited area to reduce the cost and to achieve scalability  How to ensure a quick converging construction of such a preference information?  How to achieve the global optimization with the information in those limited areas 12/7/2015Hong Kong PolyU

10. Our approach Descriptor S  [0, 1 ] ◦ Representative of preference, not ETX metric  The higher its value, a better routing path there likely will be to reach the boundary of the network  Used for routing decision to select the successor with a relatively high index value among all available neighbors  Use a single reference (path to network boundary) to reach the destination  Interchangeable use multiple references to approach to the destination  A tradeoff between cost and accuracy of information!!! ◦ S(u) = max { S(n (u) ) }  Relatively stable and quickly converging 12/7/2015Hong Kong PolyU

11. Detailed Process Network Model Information Construction ◦ Collection and distribution Information Utilization 12/7/2015Hong Kong PolyU

12. Network Model 12/7/2015Hong Kong PolyU

13. Asynchronous MAC Layer Support Faster Less synchronization overhead More accurate to describe the link status 12/7/2015Hong Kong PolyU

14. Neighbor Node Appearance The appearance of neighbor node v is determined by the Berkeley Mica mote platform as follows, with respect to the distance of link (i.e., D(u, v) = | L(u) − L(v) |). ∈ (0.9, 1], D(u, v) ≤ 10 feet ≃ 0, D(u, v) > 40 feet ∈ (0, 1), otherwise (1 u → v = 12/7/2015Hong Kong PolyU

15. Reachability Description of 1 -hop link quality Determined by the Monte Carlo method ◦ Ratio of the time that a node v appears to the total elapsed time ◦ Estimated by success REQ/ACK processes, supported by our asynchronous MAC scheme Calculated as: {v,u} ≈ u → v × v → u, 12/7/2015Hong Kong PolyU

16. Forwarding Zone and Request Zone 12/7/2015Hong Kong PolyU

17. Information Construction Initialization Phase ◦ Each node u outside the interest area sets S(u) to a fixed ( 1, 1, · · ·, 1 ); otherwise, sets S(u) to a changeable (0, 0, · · ·, 0). ◦ Then, each node will have stable status by applying S i (u) = max{ {u,v} × S i (v)}, 1 ≤ i ≤ 4 (2 and S i (u) = max{S’ i (u), {u,v} × S i (v)}, 1 ≤ i ≤ 4 (3 ◦ Such a link {u, v} is called a key link for S i (u). 12/7/2015Hong Kong PolyU

18. Identification Phase ◦ Any node u is called a type-i stuck node if it does not have any neighbor appearing inside forwarding zone Q i. Set S i (u) = 0. ◦ Uppon detecting a change of the other end of the key link, a node u with S i (u) > 0  Calculate its type-i status by using Eq. (2)  Inform all neighbors its new S i (u) in the next round  If S i (u) = 0, u is called a type-i unsafe node and no longer change its status; otherwise, u is still type-i safe and S i (u) will eventually stabilize by using Eq. (3). 12/7/2015Hong Kong PolyU

19. Self-healing phase ◦ Any node u (stuck, unsafe, or safe) will recalculate its S i (u) by using Eq. (3), until the value becomes stable. 12/7/2015Hong Kong PolyU

20. Information Utilization If d  n(u), v = d. Determine the request zone Z k (u, d) ( 1  k  4 ), according to L(u) and L(d). Select v  n(u)  Z k (u, d), where the forwarding from v to d is safe with respect to request zone Z k (v, d). 12/7/2015Hong Kong PolyU

21. Routing Properties A straightforward path can be derived when the destination d is in one type of safe area. Such a forwarding, say type-i, can be initiated at a source that has a safe successor, i.e., a type-j safe neighbor. The initiated routing may interrupt when the destination is in an unsafe area and disconnected with the source. Before the retransmission starts, the length of the path approximates to D(s, d) + , where  is the maximum length of the boundary circling an unsafe area. 12/7/2015Hong Kong PolyU

22. When s is inside an unsafe area, a successful routing will achieve a path shorter than D(s, d) +  /2. If our forwarding advances can reach the destination d with updated safety information, a path can also be constructed with outdated (or lagged) information. The self-healing phase converges in a limited number of rounds and will not affect any existing safety-information-based routing. 12/7/2015Hong Kong PolyU

23. Conclusions Traditional source routing is not applicable in highly dense and dynamic WASNs. A preference information is more suitable for forwarding routing, compared with a costly ETX like metric. Localized method to achieve global optimization in WASN is possible, but is very difficult by the consideration of overhead. With the support of MAC, a routing without synchronizing neighbors is faster and can allow more concurrent communications, enhancing the network performance. 12/7/2015Hong Kong PolyU

24. Thank you! Questions and Commons 12/7/2015Hong Kong PolyU