Title: Location awareness and localization
1Location awareness and localization
- Michael Allen
- 307CR
- allenm_at_coventry.ac.uk
Much of this lecture is based on a 213 guest
lecture on localization given at UCLA by Lewis
Girod
2Location awareness/localization?
- Where am I relative to known positions?
- Why would I want to know that?
- Where is this unknown thing relative to me?
- Why do I want to know?
3What are relevant applications?
- Navigation, tracking
- SatNav, Radar
- Target localization, monitoring
- Birds, people
- Service awareness
- Smart offices, service discovery
- Must be taken in context of application
- May be (x,y,z) coordinates (or lon, lat)
- in this room, near this device
- Can achieve this actively or passively
4Active Mechanisms
- Non-cooperative
- System emits signal, deduces target location from
distortions in signal returns - e.g. radar and reflective sonar systems
- Cooperative Target
- Target emits a signal with known characteristics
system deduces location by detecting signal - e.g. Active Bat
- Cooperative Infrastructure
- Elements of infrastructure emit signals target
deduces location from detection of signals - e.g. GPS, MIT Cricket
5Passive Mechanisms
- Passive Target Localization
- Signals normally emitted by the target are
detected (e.g. birdcall) - Several nodes detect candidate events and
cooperate to localize it by cross-correlation - Passive Self-Localization
- A single node estimates distance to a set of
beacons (e.g. 802.11 bases in RADAR) - Blind Localization
- Passive localization without a priori knowledge
of target characteristics - Acoustic blind beamforming (Yao et al.)
6Measuring success
- Simplest way is distance from ground truth
- Euclidean distance from (x,y) estimate to (x,y)
truth - Other factors
- Precision v Accuracy
- How accurate does it needto be?
- Scale
- Application requirements
High accuracy, Low precision
Low accuracy, High precision
7Measuring success II
- The less control we have over the signals we use
to estimate position, the less accuracy we can
get - Localizing a bird call is more difficult than
acoustic ToF between two nodes - No synchronisation between un-cooperative targets
- Even if we control the signals, they may have
varying degrees of accuracy - Signal strength vs acoustic/ultrasonic ranging
- Environmental problems
- Trade-off between cost, application requirements
and environment
8Ranging mechanisms
- Need some way to determine relative distances
between unknown and known positions - Timing the reception of signals that are known to
propagate at a certain speed are valuable - Audible acoustic
- Ultrasound
- Radio
- Other methods based on inverse relationship
between loss and distance - Received signal strength (RSSI)
9Time-of-Flight (ToF)
- Send two signals that propagate at different
speeds at the same time - Measure the difference in their arrival time and
use this to estimate distance - Know propagation speeds a priori
- Need to be able to detect FIRST onset of signal
- Problems
- Non-line of sight, reverb/echoes (multi-path)
- RF and acoustics are two common examples
- Radio and ultrasound
- Radio and audible acoustic
10Time-of-Flight (ToF) Example
- Radio channel is used to synchronize the sender
and receiver - Coded acoustic signal is emitted at the sender
and detected at the emitter. ToF determined by
comparing arrival of RF and acoustic signals
Radio
Radio
CPU
CPU
Speaker
Microphone
11Multipath/Non line of sight
- Multipath when signal bounces off obstacles in
the environment - Causes signal degradation for direct path
component - May estimate echoes as actual start of signal
BAD - Non line of sight when there is no direct path
between A and B - Distance A-B is now biased by some unknown
constant making it an over-estimate
A
B
12Echoes
13Ultrasonic and Acoustic ToF
- Ultrasound better suited to indoor environments
and shorter distances (10m) - Highly accurate, but highly directional
- Ultrasound less invasive
- Consider application constraints..?
- Both have multi-path and non-line of sight
problems - Echoes cause false/late detections (bias result)
- If no direction LoS, cannot ever estimate correct
range (not aware that range is incorrect!)
14RSSI/Received Signal Strength
- RSSI can be used for distance estimation
- Loss is inversely proportional to distance
covered - RSSI is bad for high accuracy
- Path loss characteristics depend on environment
(1/rn) - Shadowing depends on environment
- Potential applications
- Approximate localization of mobile nodes,
proximity determination - Database techniques (RADAR)
Path loss Shadowing Fading
Distance
15Localization primitives and examples
16Localization example - GPS
- Satellites orbit the planet, transmitting coded
signals - Atomic clocks, highly accurate
- Know own position to high accuracy
- Estimate distance through locking into coded
sequence from satellite - Our GPS devices have inaccurate clocks
- lock onto GPS signals from separate satellites
- Create local versions of the signals they are
sending - Figure out offset of our version to theirs ToF
- 3 ranges to satellites minimum reqd
- Solve problem using tri-lateration
- Accuracy of metres
17Tri-lateration/multi-lateration
- Given several known positions, and distances
from these to an unknown source, we can estimate
the position of the unknown - In 2D this is figuring out the intersection of
circles, in 3D is intersection of spheres
(slightly harder) - 3 minimum to resolve 2D ambiguity, 4 for 3D
- BUT - GPS can get away with 3 how come?
- Important primitive inposition estimation
- WSN Localization algorithmsoften built on top of
this - Multi-lateration is when you usemore than 3
- The generalisation for many observationsand 3D
18Geometry matters!!!
- If known positions are bunched together and the
unknown is far away from themGeometric Dilution
of Precision can occur - The angles relative to the unknown are too
similar, and the precision of the position
estimate is compromised - Estimate can get pushed out with poor distance
estimation - Best geometry is the convex hull (unknown is
surrounded)
GOOD
BAD
19Active bats/active badge
- ATT Cambridge (as was)
- Location system
- Badge infrared, room granularity
- Bats ultrasonic, 3D position within room
- Uses ultrasonic ranging
- Devices broadcast unique pings
- Trilateration/multilateration
- Can use same cheat as GPS
- Ceiling mounted detectors
- Centralised computation
- Device doesnt know where it is, system does
Bat
Badge
20Cricket location support system
- Similar application ideas to active bats
- Part of MIT oxygen project
- Active beacons and passive listeners
- Beacons broadcast, devices can figure out where
they are - Scales well
- Decentralised
- Low-power, reconfigurable
21Radar/Microsoft
- Uses signal strength (RSSI) to collect signature
traces of users (with laptops 802.11) - These traces can be matched to known RSSI
signatures held in a database - Position can be estimated based on comparison
- Median accuracy 2-3 metres, large variance
- Problems RSSI is not accurate, estimates will
vary even when stationary! - Expect best of 1 1.5m accuracy
- Is this good enough?
- Motetrack at Harvard did similar with motes
http//www.eecs.harvard.edu/konrad/projects/mote
track/
22Localization in a wireless sensor networking
context
- We deploy a wireless sensor network because we
want to sense and process data related to a
physical phenomena - Need to determine physical locations of sensors
to put context to data being gathered - Granularity relates to application, scale
23Goals of WSN localization
- Minimise the amount of known locations we need a
priori - Cant just give all nodes GPS.. Can we?
- Estimate ranges as cheaply as possible
- Use hardware we already have/need to use
- Maximise accuracy
- Relative to our application
- Consider scale, granularity
24Multi-hop localization
- In previous examples, devices have always been 1
logical hop away from known positions - Not necessarily the case in wireless sensor
networks - Need to design algorithms to deal with this
problem - Consider error in measurement propagates over
multiple hops - Especially bad in large networks, with poor
ranging techniques
25Case study Acoustic ENSBox
- Designed for acoustic sensing applications
- Example localizing animals based on their calls
- Passive, non-cooperative
- Highly accurate self-localization
- Acoustic ToF ranging and DoA
- Iterative multi-lateration algorithm
- Requires no a priori information
- Accuracy is important for application
- Using self-localization as ground-truth for
localizing animals - Nodes have 48KHz sampling, powerful processors,
large amount of memory
26Source-localization
- Processing chain
- Detect event (we dont control signal)
- Estimate DoA (Problem cannot rely on ToF)
- Group similar events together
- Fuse data
One node sub array All nodes array
27Results
- Ground truth is hard to define when youre
estimating non-cooperative sources! - Best hope is precision
28Conclusions
- Location awareness/localization is important
- Considered in context!!
- High accuracy can be achieved, dependent on
ranging technology, constraints of environment - Need to consider application requirements
- There are many different ranging approaches
- Approaches vary based on indoor/outdoor, size of
devices, cost, goals - Multi-hop ranging brings other challenges
- Propagation of error..