Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs)

1. Project: IEEE P802.15 Working Group for Wireless Personal Area Networks (WPANs) Submission Title: [IEEE 802.15.5 WPAN Mesh Networks] Date Submitted: [12 May, 2005] Source: [Jianliang Zheng, Yong Liu, Chunhui Zhu, Marcus Wong, Myung Lee] Company [Samsung] Address [Samsung Lab@CUNY, Steinman Hall, 140th St & Convent Ave, New York, NY 10031, USA] Voice:[+1-212-650-7260], FAX: [+1-212-650-8249], E-Mail:[lee@ccny.cuny.edu] Re: [Call for Proposal: IEEE P802.15-5/0071] Abstract: [This proposal discusses Samsung’s proposal for IEEE 802.15.5 WPAN Mesh, based on Meshed-Tree approach, including Meshed Tree routing, multicasting, and Key pre-distribution.] Purpose: [This proposal is provided to be adopted as a recommended practice for IEEE WPAN Mesh] Notice: This document has been prepared to assist the IEEE P802.15. It is offered as a basis for discussion and is not binding on the contributing individual(s) or organization(s). The material in this document is subject to change in form and content after further study. The contributor(s) reserve(s) the right to add, amend or withdraw material contained herein. Release: The contributor acknowledges and accepts that this contribution becomes the property of IEEE and may be made publicly available by P802.15. Zheng, Liu, Zhu, Wong, Lee

2. IEEE 802.15.5 WPAN Mesh Networks Jianliang Zheng, Yong Liu, Chunhui Zhu, Marcus Wong, Myung Lee Samsung Lab @ CUNY Zheng, Liu, Zhu, Wong, Lee

3. Objectives • To construct Mesh Networking Layer over IEEE 802.15.4 MAC and PHY • Proposed Mesh Network includes following features: • Meshed Tree Formation • Block Addressing • Routing • Multicasting • Key Pre-distribution • Extensible to IEEE 802.15.3 MAC and PHY Zheng, Liu, Zhu, Wong, Lee

4. Contents • Meshed Tree • Beacon-enabled Network • Tree-Based Multicasting • Key Pre-distribution Zheng, Liu, Zhu, Wong, Lee

5. MeshedTree Zheng, Liu, Zhu, Wong, Lee

6. Outline • Adaptive Robust Tree (ART) • Three phases • Meshed ART (MART) • Mesh Networking • Routing Table • Data Forwarding • Route Discovery • Route Repair • Summary Zheng, Liu, Zhu, Wong, Lee

7. 1. Adaptive Robust Tree ( ART) • Three phases • Meshed ART (MART) Zheng, Liu, Zhu, Wong, Lee

8. Three Phases Initialization Phase Normal Phase Recovery Phase Zheng, Liu, Zhu, Wong, Lee

9. Initialization Phase [children#][children#]=[8][6] A [beg,end,next]=[1,16,1][beg,end,next]=[17,28,17] • Stage 1: Association [5][2] [5] B J • Stage 2: Reporting number of children [3,12,3][13,16,13] [19,28,19] C H K [1][2][1] [3][1] [1] • Stage 3: Address assignment • An ART is formed. • Additional addresses can be reserved. [5,6,5][7,10,7][11,12,11] [21,26,21][27,28,27] [15,16,15] D E G I L O [1][1] [0] [0] [1] [0] [0] [23,24,23][25,26,25] [9,10,9] F M N [0] [0] [0] Zheng, Liu, Zhu, Wong, Lee

10. Normal Phase [8][6] A0 [1,16,1] • Normal data transmissions [17,28,17] [5][2] [5] B1 J17 [3,12,3][13,16,13] [19,28,19] • Example: Node C node L C3 H13 K19 [1][2][1] [3][1] [1] • Nodes are still allowed to join the network [5,6,5][7,10,7][11,12,11] [21,26,21] [15,16,15] [27,28,27] D5 E7 G11 I15 L21 O27 [1][1] [0] [0] [1] [0] [0] [23,24,23][25,26,25] [9,10,9] F9 M23 N25 [0] [0] [0] Zheng, Liu, Zhu, Wong, Lee

11. Recovery Phase • Tree route repair will be discussed later. Zheng, Liu, Zhu, Wong, Lee

12. Meshed ART (MART) [8][6] [1,16,1][17,28,17] A0 • Neighbors treat each other as a child. [5][2] [5] [3,12,3][13,16,13] [19,28,19] B1 J17 [1,16,1] [17,28,17] [13,16,13] • Shorter path C H13 [1] K [15,16,15] • Elimination of SPOFs [17,28,17] …… D E G I L O F M N Zheng, Liu, Zhu, Wong, Lee

13. 2. Mesh Networking • Routing Tables: • ART Table (ARTT) • Non-Tree Table (NTT) • Route Type • Data Forwarding • Route Discovery • Route Repair • Tree route repair • Non-tree route repair Zheng, Liu, Zhu, Wong, Lee

14. ART Table (ARTT) a a Zheng, Liu, Zhu, Wong, Lee

15. Non-Tree Table (NTT) Zheng, Liu, Zhu, Wong, Lee

16. Route Type • Each NTT entry provides • an optimal route to address beg_addri • an auxiliary route to the whole address block [beg_addri+1, end_addri] • In terms of cost, roughly we have: ART_route ≥ MART_route ≥ NT_auxiliary_route ≥ NT_optimal_route Whenever a route is optimal, equal sign applies from there on. Zheng, Liu, Zhu, Wong, Lee

17. Start Find an optimal route in ARTT? Y N Find an optimal route in NTT? Y N Find an auxiliary route in NTT? Y N 1 2 3 4 Use the route found Use tree route 1 2 Optimal routes 3 4 Non-optimal routes Data Forwarding Zheng, Liu, Zhu, Wong, Lee

18. Route Discovery (1) • Route Request (RREQ) and Route Reply (RREP) Packet Formats Zheng, Liu, Zhu, Wong, Lee

19. Route Discovery (2) Zheng, Liu, Zhu, Wong, Lee

20. Route Discovery (3) A • Case 1: Source has an optimal route • No route discovery B J C H K D E G I L O • Example 1:node F nodeI(optimal non-tree route) F M N • Example 2:node J nodeM(optimal tree route) Optimal non-tree route Optimal tree route Zheng, Liu, Zhu, Wong, Lee

21. A B J C H K D E G I L O F M N Route Discovery (4) • Case 2: Source has no optimal route; but destination has. dst. • Example 1:node F nodeI • Bi-directional routes are set up dst. • Example 2:node N nodeJ • No routing entry created src. src. unicast RREQ unicast RREP existing optimal non-tree route Zheng, Liu, Zhu, Wong, Lee

22. Route Discovery (5) A • Case 3: Neither the source nor the destination has optimal route. B J C H K • Example:node I nodeO D E G I L O src. dst. F M N unicast RREQ broadcast RREQ RREP Zheng, Liu, Zhu, Wong, Lee

23. Route Repair • Tree Route Repair • Non-Tree Route Repair Zheng, Liu, Zhu, Wong, Lee

24. MART route RREQ RREP Grat. RREP Tree Route Repair A • Node J broadcasts an RREQ to locate node K, with a limited TTL. • Node K fails B J C H K • All nodes below node K that have received the RREQ reply. D E G I L O • Node J selects the best path and sends a gratuitous RREP to activate it. F M N Zheng, Liu, Zhu, Wong, Lee

25. Tree Route Repair (cont.) Data Forwarding after tree route repair A [desIn,21,26,high,13] B J17 C H13 K [desIn,21,26,normal,21] [srcIn,21,26,normal,17] D E G I L21 O [1][1] [23,24,23][25,26,25] parent=13 F M N Zheng, Liu, Zhu, Wong, Lee

26. Non-Tree Route Repair • Use local repair Zheng, Liu, Zhu, Wong, Lee

27. HighLights • Adaptive address assignment • avoiding “running out of addresses” problem • Efficient tree repair • no address re-assignment • Meshed ART (MART) • shorter path • Robustness • Mesh networking (Tree routing + Non-tree routing) • optimal routes • no broadcast (even with limited TTL) if either the source or the destination has an optimal route • no flooding if there is a (non-optimal) route from the source to the destination Zheng, Liu, Zhu, Wong, Lee

28. Mesh Networking for Low Duty-cycle Networks Zheng, Liu, Zhu, Wong, Lee

29. Outlines • Basic mechanisms • Enhancements • Highlights Zheng, Liu, Zhu, Wong, Lee

30. Basic Mechanisms • Tree formation • Tree addressing • Tree routing • Topology server setup • Beacon scheduling • Reactive shortcut formation • Two-address strategy • Route repair Zheng, Liu, Zhu, Wong, Lee

31. Tree Formation • The node initiating the network becomes the PAN coordinator. • In the network formation stage, all coordinators shall enable their receivers to catch beacon requests from new nodes. • No regular beaconing is allowed before the beacon scheduling is done. • New nodes perform active scan to collect beacons from their neighbors. • Every new node selects a neighbor, which has the best path quality to the PAN coordinator, as its parent. Zheng, Liu, Zhu, Wong, Lee

32. Tree Addressing (1) • The PAN coordinator broadcasts a descendant statistics message along the tree to all leaf nodes. • Each leaf node returns a descendant counter to its parent with the initial counter value set to one. • After a coordinator collects the descendant counters from all its children, it adds them together and passes the sum to its own parent. • This process continues until the PAN coordinator receives the descendant counters from all its children. Zheng, Liu, Zhu, Wong, Lee

33. Tree Addressing (2) • The PAN coordinator divides the available address* block among its children based on the ratios of their descendant numbers. • Each coordinator, once receiving the address block assigned by its parent, further divides the address block among its own children. • This process continues until all leaf nodes receive their addresses. * The address here is MAC short address Zheng, Liu, Zhu, Wong, Lee

34. Tree Addressing (3) Zheng, Liu, Zhu, Wong, Lee

35. Tree Routing (1) • When a coordinator receives a packet not destined for itself: • If the destination’s address is out of its address block, it forwards the packet to its parent; • Otherwise, it compares the destination’s address with its children’s addresses. The packet is forwarded to child i if the destination’s address is between the addresses of child i and child i+1. Zheng, Liu, Zhu, Wong, Lee

36. Tree Routing (2) • Node H sends a packet to node F. • As H does not have any child, it forwards the packet to its parent C. • C finds that F has an address out of its address block. So it forwards the packet to its parent A. • As F’s address falls into A’s address block, and A further finds that F’s address is between the addresses of child B and C, so A forwards the packet to B. • B forwards the packet to F. Zheng, Liu, Zhu, Wong, Lee

37. Topology Server Setup • Either the PAN coordinator or a resource sufficient node can serve as the topology server. • All other nodes can reach the topology server by using tree routing. • Each coordinator shall report its superframe parameters and link states to the topology server. • Each coordinator may periodically scan its neighbors' beacons and report significant link changes to the server. • There can be two or more topology servers acting as backup of each other. Zheng, Liu, Zhu, Wong, Lee

38. Beacon Scheduling • When receiving the neighboring information of a coordinator, the topology server assigns a contention-free beacon time-slot to the coordinator. • Every coordinator gets a beacon time slot that is not overlapped with the active periods of its two-hop neighbors. • This two-hop beacon scheduling ensures that each node can correctly capture all its neighbors’ beacons and locate their active periods. • Once a coordinator receives the beacon time assignment, it can emit regular beacons and operate in beacon-enabled mode. Zheng, Liu, Zhu, Wong, Lee

39. Reactive Shortcut Formation • An active source sends a shortcut request (SCRQ) message to the topology server. • The topology server calculates the optimal shortcut by using the Dijkstra’s algorithm. • The topology server sends a shortcut notification (SCNF) message to the destination. • The destination sends a shortcut reply (SCRP) message to the source to establish routing entries along the shortcut. • Relay nodes along the shortcut shall locate and record the active periods of their previous-hop and next-hop neighbors. Zheng, Liu, Zhu, Wong, Lee

40. Two-address Strategy • Two addresses are used in routing: • MAC short address is used in tree routing. • NWK address is used in shortcut routing. • MAC short addresses may be changed with the variations of the tree structure. • NWK addresses shall not be changed. • Each packet shall carry the destination’s NWK address. When the packet is delivered through a tree route, it shall also carry the destination’s MAC short address. Zheng, Liu, Zhu, Wong, Lee

41. Route Repair (1) • If the topology server is not the PAN coordinator, it keeps maintaining the route between it and the PAN coordinator. • When a node detects the failure of a non-tree link or a tree link to its child, • It updates the link change to the topology server. • If it cannot reach the topology server directly, it may ask the PAN coordinator to forward the link change. • The topology server recalculates a new route between this node and the destination. Zheng, Liu, Zhu, Wong, Lee

42. Route Repair (2) • When a node cannot reach its parent, • It initiates tree reconstruction by dismissing all its descendants. • All affected nodes rejoin the tree and obtain new MAC short addresses (NWK addresses are not changed). • The node detecting the link failure updates the topology server about the link change. • If there is an address mapping server, the nodes with new MAC short addresses shall update the server. • The sources with the old addresses of desired destinations either contact the address server or broadcast an address search message along the tree to obtain the new addresses. • The existing shortcuts are not affected since they depend on NWK addresses. Zheng, Liu, Zhu, Wong, Lee

43. Enhancements • Solutions to potential beacon collisions • Solutions to single point of failure • Solutions to large network case Zheng, Liu, Zhu, Wong, Lee

44. Solutions to Beacon Collisions (1) • When node N does not appear in the network, node C and E are not two-hop neighbors. • The topology server may assign the same beacon time slot to C and E. • When node N presents and tries to join the tree, it cannot receive beacons from either C or E. Zheng, Liu, Zhu, Wong, Lee

45. Solutions to Beacon Collisions (2) • Solutions: • New nodes may scan not only beacons, but also other packets. According to the scanning results, the new nodes may estimate active periods of their neighbors, and request the neighbors to send ad hoc beacons. • The topology server may define an irregular beacon period that is not overlapped with the active period of any coordinator, so that each coordinator can send irregular beacons during this period in randomly selected superframes. Zheng, Liu, Zhu, Wong, Lee

46. Solutions to SPF (1) • Backup server • The main server shall synchronize its topology knowledge with the backup server. • The neighbors of the main server record the address of the backup server. • Once the main server fails, all messages destined for the main server are forwarded to the backup server. Zheng, Liu, Zhu, Wong, Lee

47. Solutions to SPF (2) • When the PAN coordinator acts as the topology server, • The neighbors of the PAN coordinator may establish alternative table-driven routes to reach each other (bypassing the PAN coordinator). • Once the PAN coordinator fails, its neighbors can still maintain the tree routing. Zheng, Liu, Zhu, Wong, Lee

48. Solutions to SPF (3) • The PAN coordinator R has three children A, B, and C. • Alternative routes are established among these children. • When R fails, node A can forward packets to C’s descendants through an alternative route A-B-C. Zheng, Liu, Zhu, Wong, Lee

49. Solutions to Large Network Case (1) • If a topology server cannot accommodate the whole network topology, it may record only the top portion of the tree. • The topology server can help nodes within the server coverage area (SCA) establish optimal shortcuts to reach each other. • Nodes outside the SCA have to use tree routes. • If a tree route passes through the SCA, the topology server can optimize the part of the tree route lying within the SCA. Partial Topology Server Zheng, Liu, Zhu, Wong, Lee

50. Solutions to Large Network Case (2) • If multiple nodes can act as topology servers, the network can be partitioned into several subtrees, and each subtree is assigned a local topology server. • Communications within a subtree can be handled by its local topology server. • A central topology server is also named to manage the communications among different subtrees. • The central topology server can help the roots of different subtrees establish shortcuts to connect each other. Zheng, Liu, Zhu, Wong, Lee