520 likes | 687 Views
Position-based Routing in Ad Hoc Networks. Brad Stephenson A presentation submitted in partial fulfillment of the requirements of the course ECSE 6962. Objectives. Introduction to position-based routing Discuss location services Discuss specific routing algorithms Greedy algorithm
E N D
Position-based Routing inAd Hoc Networks Brad Stephenson A presentation submitted in partial fulfillment of the requirements of the course ECSE 6962
Objectives • Introduction to position-based routing • Discuss location services • Discuss specific routing algorithms • Greedy algorithm • Directional flooding algorithm • Hierarchical algorithm • Comparison with topology-based algorithms
Review • Topology-based routing • Uses information about the (virtual) links that exist in a wireless network • Can be: • Proactive • Reactive • Hybrid
Position-based Routing • Additional information is used to make routing decisions, namely the physical location of the node • Decisions made based on destination’s position and position of forwarding node’s neighbors • Uses a location service to obtain the location of the destination node
Position-based Routing • Does not require routing tables • Traffic overhead may be small • Supports delivery of packets to a geographical area, called geocasting [NI] • Three broad categories: • Greedy forwarding • Restricted directional flooding • Hierarchical methods
Location Services • Centralized location service • Mobile nodes register their position with the location service • The service is contacted when a routing node wishes to find a destination node • Similar to cellular network • Requires that position servers be well-known • Only works with a non-ad-hoc external service
Location Services • Decentralized location services can be: • All-for-all • All-for-some • Some-for-all • Some-for-some • See [MWH]
Decentralized Location Services DREAM [B] D Node A wants to send an update B E A F C G
Decentralized Location Services DREAM [B] D B E Node A wants to send an update A F C G
Decentralized Location Services DREAM [B] D B E Node A wants to send an update A F C G
Decentralized Location Services DREAM [B] D Spatial Resolution B E Node A wants to send an update A F C G
Decentralized Location Services DREAM [B] D Temporal Resolution B E A F C G
Decentralized Location Services Quorum-Based [MWH] I A L K D E 3 H 2 G 1 B S J C The backbone must be set up using a non-positionbased ad hoc routing mechanism
Decentralized Location Services Homezone [MWH] • Location information for node A is stored in a virtual homezone • The position of the homezone can be found by applying a well-known hash function to the node ID
Decentralized Location Services Homezone [MWH] D E F P A G C B
Key Assumptions • Unit Disk Graph (UDG) model of physical layer • Nodes are in two dimensional space • Homogeneous nodes in the network • What major limitations do these assumptions expose? • Depends on the application
Key ideas inPosition-based Routing Algorithms [GSB] • Loop-freedom • Distributed operation • Path strategy • Metrics • Memorization • Guaranteed delivery • Scalability • Robustness
Loop-freedom • Should be inherently loop-free • Avoids recovery strategies • timeout of old packets • memorizing packets that have been seen before
Distributed operation • Localized algorithms are preferred if performance matches global algorithms • Decisions made based on local information • Reduced overhead • If using n-hop neighbors, can be classified as 2-localized, 3-localized, etc.
Path Strategy • Single path • Flooding • Directional Flooding • Multipath
Metrics • Hop count • Hop quality • Power consumption • Policy-based cost • Expected hop count (accounts for retransmissions) [S02]
Memorization • Better to avoid memorizing traffic because of queue size and changes in mobility • Required for QoS-guaranteed paths
Guaranteed Delivery • Delivery rate = # delivered / # sent • Guaranteed delivery has delivery rate = 1 • To achieve this, we need a MAC protocol which provides retransmit or no collisions
Scalability • Increase in overhead as number of nodes increases • Sometimes a subjective measure
Robustness • How does mobility affect the algorithm • How accurately can we determine the position of the destination
Greedy Algorithms • Loop free [SL] • Localized information • Single path strategy • Metric: Hop count • No memory • No guarantee of delivery • Scalable, O( sqrt(n) ) [MWH] • Somewhat robust
Greedy Packet Forwarding “Send to (10, 3)” 2 4 S 3 D 1 (x, y) = (10, 3) 5 R
Greedy Packet Forwarding Most Forward within R [TK] 2 4 S 3 D 1 5 R
Greedy Packet Forwarding Nearest with Forward Progress [MWR] 2 4 S 3 D 1 5 R
Greedy Packet Forwarding Compass Routing [MWR] 2 4 S 3 D 1 5 R
Greedy Algorithms • Most forward within R • Get as far as you can within sender’s range • Nearest with forward progress • Makes collisions less likely • Compass Routing • Send to nearest neighbor that is directly between sender and receiver
Greedy Routing Failure [MWH] Local maximum
Recovery Algorithms • Greedy Perimeter Stateless Routing Protocol (GPSR) • Face-2 algorithm • Other variants/combinations • Based on traversal of planar graphs • Returns to greedy mode when closer to destination than when it entered recovery
Recovery Algorithms • Construct the planar subgraph [T] • Forward the packet along interior face using the right hand rule
Recovery Algorithms [MWH]
Recovery Algorithms D 4 3 5 2 Scan begins at incoming edge Assume communication only occurs along the edges of the planar graph S 1
Recovery Algorithms D 4 3 5 2 Recovery complete! Revert back to greedy mode Assume communication only occurs along the edges of the planar graph S 1
Restricted Directional Flooding • Not loop free • Localized operation • Path strategy: flooding/multipath • Metric: Hop count • Memory • No guarantee of delivery • Not scalable, O(n) [MWH] • Not robust
Restricted Directional Flooding • DREAM and LAR • Send packet to all neighbors “in the direction” of D • How do we determine this direction?
Restricted Directional Flooding DREAM Expected Region [B] S q R D Expected Region
Restricted Directional Flooding DREAM Expected Region [B] • Needs a recovery mechanism if no neighbor is in the direction of the expected region • None specified in DREAM proposal • Area of future work
Restricted Directional Flooding Location-Aided Routing [KV] • Uses the idea of restricted flooding toward the expected region for path discovery in non-position-based routing protocols [KV]
Hierarchical Routing • Terminodes and Grid Routing • Possibly reduces the complexity of information each node has to handle • Improves scalability • Can ad hoc networks also reap these benefits? • Not without tradeoffs!
Hierarchical Routing Grid Routing [MWH] • Uses greedy approach for long-distance routing • Uses non-position-based approach at the local level (proactive distance vector) • Allows non-position-aware nodes to participate • More tolerant of position inaccuracy • More complex to implement
Topological vs. Positional • Terminodes shown to improve packet delivery rates and overhead compared to reactive ad hoc routing [BGL] • GPSR performs better than DSR in almost all criteria including overhead and delivery rate [Br] • Both results are from simulations
Are there any applications? • Vehicle-to-vehicle communication networks • Geocasting can be useful for … • Tactical military information • Disaster response • Personalized Internet experience • Home security
(IMHO) • Very little experimental work done, mostly simulation • Assumptions limit the scope, practicality of results • Solution: Need more engineering graduate students to conduct experiments
Future Work • There is a plethora of ideas • Quantitative work must be performed • Investigate hashing in highly dynamic networks • Probabilistic approach • Recovery strategies within constraints • Deeper hierarchies (3-tier, etc.) • What about anonymity?
Open Problems Remaining • Mobility-caused loops • Congestion considerations (replace hop count metric with e2e delay) • Quality of Service considerations • An excellent recent paper on using a non-UDG model is [SNK]