Presentation on theme: "Range-Based and Range-Free Localization Schemes for Sensor Networks."— Presentation transcript:
Range-Based and Range-Free Localization Schemes for Sensor Networks
Localization Critical service A sensor reading consists of E.g., target tracking, disaster recovery, fire detection, patient location in a smart hospital, … Needed for geographic routing Too expensive for an individual sensor to have a GPS (Global Positioning System) Reference nodes (called anchor or beacon nodes) + sensor nodes
Range-based localization schemes TOA (Time of Arrival) Get range info via signal propagation delay E.g., GPS Expensive, power consuming, inaccurate TDOA (Time Difference of Arrival) Transmit both radio and ultrasonic signals at the same time to observe the arrival time difference Extra hardware, i.e., ultrasonic channel, is required Not only radio but also sound signals have multipath effects affected by humidity, temperature, …
Received signal strength (RSS) Distance estimation based on RSS Hard due to radio signal vagaries AoA (Angle of Arrival) A node estimates the relative angles between neighbors Requires directional antennae
Range-free localization Centroid algorithm Anchors beacon their positions to neighbors (single hop broadcast) A sensor node computes the centroid using all received beacon messages
DV-HOP Anchor locations are flooded through the network Keep the running hop count Estimate average one hop distance Amorphous Positioning Similar to DV-HOP Use offline one hop distance estimation
Range-Free Localization Schmes for Large Scale Sensor NEtworks - APIT (Approximate Point In Triangulation) Mobicom 2003
PIT (Point In Triangulation) A node chooses three anchors from all audible anchors Test whether it’s inside the triangle Repeat for all possible combinations of audible three anchors Compute the COG of the intersection of all the triangles
Perfect PIT test For three given anchors, A, B, C, determine whether a point M with an unknown position is inside the triangle ABC or not Proposition I: If M is inside the triangle, when M is shifted, the new position is nearer to (or farther from) at least one anchor A, B, or C
Proposition II: If M is outside the triangle, when M is shifted, there must exist a direction in which the position of M is farther from or closer to all three anchors A, B and C
Problems with Perfect PIT test How can a sensor node perform the PIT test w/o actually moving? How to do exhaustive tests considering all possible directions of departure?
APIT (Approximate PIT test) In a certain propagation direction, the received signal strength is assumed to monotonically decrease in an environment w/o obstacles Departure test
Signal strength at different distances (to justify the departure test)
APIT test Basic idea: Use neighbor info, exchanged via beaconing, to emulate the node movement in the perfect PIT test If no neighbor of M is farther from/closer to all three anchors A, B & C simultaneously, M assumes that it is inside the triangle.
Errors in the APIT test InToOut Error OutToIn Error
APIT error measurements 14% error when a node has 6 one-hop neighbors in average – Small?
APIT aggregation: Mask errors in individual APIT tests Aggregate individual APIT test results through a grid SCAN Length of a grid side is 0.1R For each inside decision, the values of the grid regions over which the triangle resides are incremented Decrement for each outside decision Find the area with max values Take the center of gravity for position estimation
APIT algorithm 1. Each node maintains a table of anchor ID, location & signal strength
2. Nodes exchange anchor tables with the neighbors
3. Run the PIT test for each column of the table 4. Repeat step 3 for varying combinations of three anchors 5. Use the APIT aggregation alg. to determine the area w/ max overlap 6. Final location estimation = COG of that area
Performance evaluation Radio model Upper & lower bounds on signal strength Beyond the UB, all nodes are out of communication range Within the LB, every node is within the comm. range Between LB & UB, there is (1) symmetric communication, (2) unidirectional comm., or (3) no comm. Degree of irregularity (DOI)
Simulation parameters Node density (ND) Anchors heard (AH) Anchor to node range ratio (ANR) Avrg distance an anchor beacon travels/avrg distance a regular node signal travels Anchor percentage (AP) DOI GPS error Placement: uniform or random
Localization error for varying AH APIT works better as AH increases. Large errors when AH < 8 It’s relatively less sensitive to random deployment.
Localization error for varying ND Amorphous has large errors when ND < 10 APIT & DV-Hop show good perf if ND >= 6 Amorphous is more sensitive to larger DOI
Localization error for varying ANR Error increases as ANR increases due to error accumulations APIT has large errors when ANR < 3 due to large InToOut error
Localization error for varying DOI Irregular hop count distribution in Amorphous & DV-Hop
Communication overhead for varied AH Amorphous & DV-Hop rely on the flooding of anchor beacons
Communication overhead for varied ND
Localization error impact on geographic forwarding
Summary APIT is resilient to irregular radio patterns and random deployment Relatively low overhead compared to DV-Hop & Amorphous localization (but more overhead than Centroid) Localization has been well studied but still needs more work
What is location verification? Different assumptions from general localization What if some malicious nodes lie about their lcoation? Sample attack scenario Cliam to be very close to the sink Attract many packets Drop some or all of them Very easy DoS attack especially for geographic routing protocols
How SerLoc works Node i claims its location is (x, y) Node i needs to send (x, y) a location verification request msg to a nearby verifier A verifier can be a normal sensor node The verifier sends a random nonce to node i and start the clock Node i has to immediately return the challenge through both radio and ultrasonic channels The verifier measures the time for node i returning the challenge and take the difference between the radio & ultrasonic signal propagation. Based on this observation, verify the claimed location
Weakness of SerLoc Requires extra hardware, i.e., ultrasonic channel Innocent victims may respond late due to backlog Not location verification but range verification Verifier M’s Real Location M’s claimed Location sink Oops... Verifier cannot tell the difference! Big trouble...
Possible Research Issues Most localization work is mathematical and evaluated via (high level) simulations More realistic work is needed Indoor localization is harder Look at CodeBlue project at Harvard Location verification Can’t trust sensors Secure localization Can’t trust anchors