View
216
Download
0
Tags:
Embed Size (px)
Citation preview
Trajectory-Based Forwarding Mechanisms for Ad-Hoc Sensor
Networks
Murat Yuksel, Ritesh Pradhan, Shivkumar KalyanaramanElectrical, Computer, and Systems Engineering Department
Rensselaer Polytechnic Institute, Troy, [email protected], [email protected], [email protected]
Rensselaer Polytechnic Institute, Troy, NY 2
Outline Motivation Overview of Trajectory-Based Routing (TBR) Bezier curves for TBR Forwarding algorithms for TBR Long trajectories Simulation results Future work
Rensselaer Polytechnic Institute, Troy, NY 3
MotivationThere may be several cases where shortest-path routing is not suitable for the application:To measure some parameters for a riverTo obtain terrain knowledge of a hostile areaTo use safer locations for important data transmissions
Such application-specific requirements are particularly important for sensor networks
Rensselaer Polytechnic Institute, Troy, NY 4
Motivation (cont’d)
Example: Consider a battlefield with east-side of mountains being friendly area.Application can request to:
obtain view of “west-side of the mountains”
transmit secure information to allied soldiers through “east-side of the mountains”
Rensselaer Polytechnic Institute, Troy, NY 5
Overview of TBRSource Routing (SR) Source inserts entire route into each packet, e.g. SBR,
DSR. Very flexible for applications, but causes too large packet
headers.Greedy Routing (GR) Assuming a positioning service, each packet is forwarded
to the neighbor closest to the destination, e.g. GPSR, CR. Fixed-size, short packet headers, but not flexible for
applications.Trajectory-Based Routing (TBR) Proposed by Nath and Niculescu from Rutgers University. Represents the whole path as a parametric curve and
encodes it into each packet. Geographic routing protocol, and requires positioning
service.
Rensselaer Polytechnic Institute, Troy, NY 6
Overview of TBR (cont’d)What happens when a packet travels in the network? Source encodes the trajectory into the packet’s header. All nodes forward the packet based on a predefined forwarding strategy. After packet arrival, the intermediate nodes decode the trajectory and
forwards the packet along the trajectory. The packet gets forwarded until it reaches the destination or is dropped.
TBR is a middle-ground between SR and GR. Since a parametric curve can form any path (e.g. circle, straight line,
oscillatory lines), it gives more flexibility to define the path. – similar to SR
Since nodes decode the trajectory, i.e. stateless – similar to GR
One important issue is “how should we encode the trajectory into packets’ headers”?
Rensselaer Polytechnic Institute, Troy, NY 7
source
destination
Control pt -2
Control pt -1
Bezier Curves for TBRWe propose to encode paths by using Bezier curves.Cubic Bezier curves (2 control pts + source + destination) are easy to handle.A Cubic Bezier curve is represented in parametric form:
Q(0) is the source point, and Q(1) is the destination point.
Rensselaer Polytechnic Institute, Troy, NY 8
Bezier Curves for TBR (cont’d)
If (x0,y0), (x1,y1), (x2,y2) and (x3,y3) are known, then the
constant vectors A, B & C can be calculated as:
Each packet header contains locations of source (x0,y0), destination (x3,y3) and control points (x1,y1), (x2,y2).
So, when a packet arrives, each node: Decodes the trajectory by performing the above
calculations Figures out which neighbor to forward the packet, based on
forwarding strategy.
Rensselaer Polytechnic Institute, Troy, NY 9
Forwarding Algorithms for TBR
Terminology: di = closest distance of node Ni to the trajectory curve ti = value of the time parameter at the point where node Ni is closest to the curve – residual of
node Ni
The residual ti of node Ni can also be interpreted as projection of the node on the curve.neighbor of Ni = set of nodes that are in transmission range of Ni and have a residual greater than ti.
Rensselaer Polytechnic Institute, Troy, NY 10
Forwarding Algorithms for TBR (cont’d)
Closest-To-Curve (CTC) - node forwards to its neighbor closest to the curve.
Least Advancement on Curve (LAC) – node forwards to its neighbor with least advancement on the curve.
Random - node randomly forwards to one of its neighbor
Rensselaer Polytechnic Institute, Troy, NY 11
Forwarding Algorithms for TBR (cont’d)
CTC-LAC – node forwards to its neighbor with LAC but is also close to the curve (within a predefined distance).Most Advancement on Curve (MAC) – node forwards to its neighbor which is nearest to the destination.
Failure of CTC and MAC Failure of LAC
Rensselaer Polytechnic Institute, Troy, NY 12
Forwarding Algorithms for TBR (cont’d)
Lowest Deviation from Curve (LDC) – node forwards to its neighbor with lowest deviation from curve.
Calculation of areas is computationally intensive.
Can be approximated by numerical techniques.
Rensselaer Polytechnic Institute, Troy, NY 13
Long Trajectories For a generalized long trajectoryWe brake the trajectory into multiple cubic Bezier curves.
Before data traffic, source performs signaling and sends a probe packet that include all the control points (more than two) for the trajectory and starting locations of the smaller cubic Bezier curves (i.e. Intermediate Point (IP) ).
Nodes close to an IP will contend for being a Special Intermediate Node (SIN).
Rensselaer Polytechnic Institute, Troy, NY 14
Long Trajectories (cont’d)SINs (i.e. I1, I2 below) do special forwarding. They remove info about last curve’s control points and
replaces it with that of the next piece’s control points from packet’s header and inserts the next one’s control points.
Rest of the nodes fwd packets to nodes that are closest to curve and you advance least on curve.
Curve 3
DCurve 1
Curve 2
S
I1
I2
Rensselaer Polytechnic Institute, Troy, NY 15
Simulation ResultsUsed NS-2Number of nodes – 50, 100, 150, 200.Area – 250mX500mThree different trajectories:
Circular Zigzag –
Single-piece Zigzag --
Multi-piece
No mobility yet
Rensselaer Polytechnic Institute, Troy, NY 16
Simulation Results (cont’d)A long trajectory composed of two concatenated cubic Bezier curves
Rensselaer Polytechnic Institute, Troy, NY 17
Simulation Results (cont’d)
Deviation of various forwarding strategies from the circular trajectory
Rensselaer Polytechnic Institute, Troy, NY 18
Simulation Results (cont’d)
Normalized path length in various forwarding strategies applied on the circular trajectory
Rensselaer Polytechnic Institute, Troy, NY 19
Simulation Results (cont’d)
Deviation of various forwarding strategies from the single-piece zigzag trajectory
Rensselaer Polytechnic Institute, Troy, NY 20
Simulation Results (cont’d) Normalized path length in various forwarding
strategies applied on the single-piece zigzag trajectory
Rensselaer Polytechnic Institute, Troy, NY 21
Simulation Results (cont’d)
Deviation from the trajectory and normalized path length for the multi-piece zigzag trajectory with CTC-LAC forwarding strategy
Rensselaer Polytechnic Institute, Troy, NY 22
Future Work Extensive simulation of multi-piece caseAmount of state maintained at SINsStrategies for selecting SINs
Simulation with various mobility patterns Analysis of success rate (i.e. % reaching destination) for the forwarding strategies Resilience strategies to increase success rate
Rensselaer Polytechnic Institute, Troy, NY 23
Thank you !!