Fundamentals of Positioning and

Navigation
12.2.2025

Prof. Heidi Kuusniemi

Professor in Computer Science,
Positioning technologies and systems - University of Vaasa,
Research Professor (part-time), Finnish Geospatial Research Institute

Navigation Technologies for Autonomous Systems

Content

• Definitions and basic concepts of positioning

and navigation

• Coordinate systems

• Time and timing systems

 Definitions
and basic
concepts of positioning &
navigation
 Definitions (1)

▸ absolute vs. relative

▸ Absolute navigation is the method of determining where you are and where you
are going using precise geospatial coordinates
▸ every object has a geographic location on Earth’s surface, and those location
coordinates are used to navigate between different places and objects
▸ Relative navigationinvolves using perception of immediate surroundings to get
from one place to another, and is typically best for obstacle avoidance

4
 Definitions (2)

▸ global vs. local positioning

▸ Global: positioning technologies designed
to operate globally, in a global reference frame
▸ Local: positioning technologiesdesigned to
operate within a limited geographical area

5
 Definitions (3) Source: PennState, GPS and GNSS for Geospatial Professionals

▸ precision vs. accuracy

6
 Position estimation methods

dynamic
static

• System has memory

• System has no memory • Predict current state
• Estimate position only from previous state
from current • Also estimate current
measurements state from current
(“snapshot”) measurements
• Example technique: • Combine the two to
Least squares minimize error
estimation • Example technique:
Kalman filter
7
Static positioning (1)

• Positioning means the estimation of a receiver’s coordinates and possibly

other interesting quantities such as velocity, orientation and clock bias

• Many kinds of measurements coming from different sources can be used,

and to solve the problem we first form a mathematical mode to describe
the measurements, and then apply suitable mathematical “machinery” to
the model

• No model describes reality perfectly

• Often a simpler model is needed to be used

• Static positioning means computing a single location estimate from

several simultaneously taken measurements
• Independent of previous or future measurements, or estimates

Source: Ali-Löytty, Collin, Sirola (2010). Mathematics for Positioning, TUT.

Static positioning (2)

• Typically, the static positioning problem can be formulated as solving for a

nonlinear system of equations either in the sense of least squares, in
closed form or with respect to likelihoods or probability • Solving a nonlinear system
• All the measured values are collected into a vector y and the system of of equations
equations is written in the form
• Seeking maximum
• y = h(x) + v likelihood
• where the measurement function h is a known vector valued
• Seeking maximum or mean
function, x is the location, v is the measurement error vector, of the Bayesian posterior
whose probability distribution is assumed known

• When measurements have been made and the measurement equations

known, the location/state estimate x̂ is searched for that fits the
measurements best

Source: Ali-Löytty, Collin, Sirola (2010). Mathematics for Positioning, TUT.

Measurement equations

• Example 1: Range measurements to a station

located at s can be written as
• r = ‖s-x‖ + v
• where r is the measured range, x is
location and ‖s-x‖ (= h(x)) is the true
range

• Example 2: A GNSS pseudorange measurement

can be written as
• r = ‖s-x‖ + b + v Positioning with pseudoranges – clock bias
• Where b is the additional error in meters shows up as an error with same magnitude
in all pseudoranges
caused by the receiver’s clock bias

Source: Ali-Löytty, Collin, Sirola (2010). Mathematics for Positioning, TUT.

Overview of estimating PVT with GNSS
 Simplifying the total receiver clock error term to b = ctu
we obtain the pseudorange equation as:
k k ~
r = x − xu + b +  Tk
c

 which relates the corrected pseudorange

measurements for the kth satellite in view to the
WGS84 position coordinates for the user-receiver
vector, the user receiver clock offset, and the
pseudorange error term
 the error term can be minimized to some extent using
various techniques but never completely eliminated
 Linear Position Estimation
• We can solve the nonlinear equations
r ck = x k − x u + b + ~Tk

using closed-form solutions, iterative techniques

based on linearization or Kalman filtering
• The standard approach for estimating the receiver
position and clock offset is first to linearize the range
measurements around a rough guess of the receiver
position and clock bias and then to iterate until the
difference between the guess and the measurements
approaches zero
• A minimum of four measurements are needed to
estimate the user receiver position and clock offset
• The rough guess of the receiver position can be very
coarse, e.g. the centre of the Earth is adequate with
GNSS
 Closed form solutions

• A closed form solution, or a direct solution, is a general name for

various non-iterative algorithms where an initial guess for the state is
not needed to start the iteration
• A range measurement defines a spherical surface in three
dimensional location space
• A range difference measurement defines a hyperboloid

• Closed form solutions are useful as an initial guess for iterative

methods
• Difficult to account for measurement errors with them
Time difference of arrival

Source: Ali-Löytty, Collin, Sirola (2010). Mathematics for Positioning, TUT.

Iterative least squares

• A residual describes the incompatibility between measurements and

an assumed location x
• p(x) = h(x)-y
• If measurements are error-free, then the residual goes to zero
when x is the true location

• The idea of the least squares methods is to determine a location

estimate x̂ that minimizes the expression ||p||2

• Usually the measurements contain some errors and the location

estimate is obtained by solving the minimization problem
• x̂ = arg minx p(x)Tp(x)

Source: Ali-Löytty, Collin, Sirola (2010). Mathematics for Positioning, TUT.

 Linear least squares (1)

▸ Suppose we want to solve the linear system of equations Ax=b when we

have more equations than unknowns (A has more columns than rows).
Ideal Noisy

15
 Linear least squares (2)

▸ LeastSquares solution: find best possible compromise, when an exact

solution doesn't exist, that is, find the solution that minimizes the
residual error ||Ax-b||
▸ xls = (ATA)-1ATb

16
 Nonlinear least squares (1)

▸ Unfortunately, most problems in real life are nonlinear

▸ e.g. GNSS trilateration r k

c = x k
− x u + b + 
~k
T

▸ Nonlinear Least Squares: start with initial guess [x0 , b] and compute
Taylor expansion r ck = f ( xu , yu , zu , b) = f ( x0 + x, y0 + y, z0 + z , b0 + b) =
f ( x0 , y0 , z0 , b0 ) f ( x0 , y0 , z0 , b0 )
f ( x0 , y0 , z0 , b0 ) + x + y +
x0 y0
f ( x0 , y0 , z0 , b0 ) f ( x0 , y0 , z0 , b0 ) 1 2 f
z + b + + ...
z0 b0 2! x 2

17
 Nonlinear least squares (2)
 With a Taylor series expansion about the
approximate position and estimated receiver clock
offset [ x0 , y0 , z0 ] and b0 , we get
r k = r ck − r 0k
f ( x0 , y0 , z0 , b0 ) f ( x0 , y0 , z0 , b0 )
= x + y +
Line-of-sight unit vector x0 y0
between the estimated
receiver location and f ( x0 , y0 , z0 , b0 ) f ( x0 , y0 , z0 , b0 )
satellite k z + b
z0 b0
Lkunit
xk − x0
=− k  x + b + ~Tk
x − x0
Nonlinear least squares (3)
 When measurements from more than one satellite
are available (n), the equation can be written
compactly as Geometry matrix G

r  (− L ) 1
1 1
unit
T
 T 
~ 1

 2   ~ 2 
 r   ( − 2
L ) T
1 x   T 
ρ = = unit
  +
     1 b    
 n  n T  ~ n 
r  (− Lunit ) 1  T 
x ~
ρ = G   + εT
b
Nonlinear least squares (4)
 The errors in the error vector ~εT are assumed to
be zero-mean and uncorrelated
 Thus, the estimate in the least squares sense for
x andb is given by x
xˆ =   = (G T G ) −1 G T ρ
b
 Least squares optimization when solution over-
determined (n>4)
 Least squares originally formulated by Gauss
 x u = x 0 + x = x 0,new
b = b0 + b = b0,new
Typically set to zero
Filtering

• A filter uses all the earlier measurements in addition to the current

ones for computing the current state estimate

• A basic prerequisite for the use of earlier measurements in positioning

is a dynamic state model, that describes how the current state
depends on the previous one
• xk = fk-1(xk-1)+wk-1
• State xk can include velocities, clock biases or other interesting
quantities in addition to location
• fk is the state transfer function and wk is the state model error

• The measurement equations are the same as in the static case

• yk = hk(xk) + vk

Source: Ali-Löytty, Collin, Sirola (2010). Mathematics for Positioning, TUT.

Kalman filtering for position computation - intro
 Computing the conditional state distribution in a time series, from which we
obtain the desired estimate, is called filtering
 The Kalman Filter (KF) is an estimator for estimating the instantaneous ‘state’
of a linear dynamic system perturbed by white noise by using measurements
linearly related to the state but corrupted by white noise
 The resulting estimator is statistically optimal with respect to any quadratic function of
estimation error
Source: M. S. Grewal, A. P. Andrews, Kalman Filtering,
Theory and Practice Using MATLAB, New York, Wiley, 2001

 Kalman filter is iterative, robust and less complex

 It reconstructs all the state variables of a linear dynamic system from noisy
observations
 KF was used in the Apollo missions to the moon
 If both the state model and the measurement
equations are linear and the errors and the initial
Kalman filtering (1) state are normally distributed, then the state
conditional distribution can be computed
recursively = Kalman filter

▸ the Kalman Filter is “the best known filter, a simple and elegant algorithm, as
an optimal recursive Bayesian estimator for a somewhat restricted class of
linear Gaussian problems” (B. Ristic et al., Beyond the Kalman filter, particle filters for tracking applications, 2004)
▸ Offers means to estimate the state 𝑥 ∈ ℛ 𝑛 of a discrete-time process with a
difference equation process noise
State transition model p(w)~N(0,Q)

𝒙𝒌 = 𝑨𝒙𝒌−𝟏 + 𝑩𝒖𝒌−𝟏 + 𝒘𝒌−𝟏

Optional control input
using measurement 𝑦 ∈ ℛ 𝑚
𝒚𝒌 = 𝑯𝒙𝒌 + 𝒗𝒌
measurement noise
p(v)~N(0,R)
Observation model
Kalman filtering (2)
▸ Feedback control:

time update measurement update

”predict” ”correct”

P = estimate
covariance
xˆ k− = Axˆ k −1 Q = process
covariance
▸ Time update: Pk− = APk −1 AT + Q
𝐾𝑘 = 𝑃𝑘− 𝐻 𝑇 (𝐻𝑃𝑘− 𝐻 𝑇 + 𝑅)−1
▸ Measurement update: 𝑥ො𝑘 = 𝑥ො𝑘− + 𝐾𝑘 (𝑦𝑘 − 𝐻 𝑥ො𝑘− )
R = measurement 𝑃𝑘 = (𝐼 − 𝐾𝑘 𝐻)𝑃𝑘−
covariance

K = Kalman gain
Kalman filtering for position computation (cont.)
 The estimate of the PVT can also be obtained by using an extended
Kalman filter (EKF)
 Is an implementation of a Bayes estimator
 Basically, a weighted, recursive least squares estimator
 The outputs from an EKF will often be better than those from the least squares
method
 EKF assumes some knowledge of the receiver dynamics
 When these assumptions are incorrect, problems can arise quickly
 Basic concept behind the EKF filter is that some of the parameters
being estimated are random processes and as data are added to the
filter, the parameter estimates depend on new data and the changes in
the process noise between measurements
Kalman filtering for position computation (cont.)
 Extended Kalman filter to solve the position:
 Measurements rt with noise t , and a state vector xt with specified statistical
properties (process noise wt)

𝛿𝜌𝑡 = 𝐻𝑡 𝛿𝐱 𝑡 + 𝜀𝑡 Observation equation at time t

𝛿𝐱 𝑡 = 𝐒𝑡 𝛿𝐱 𝑡−1 + 𝐰𝑡 State transition equation
< 𝜀𝑡 𝜀𝑡𝑇 >= 𝐑 𝑡 < 𝐰𝑡 𝐰𝑡𝑇 >= 𝐐𝑡 Covariance matrices

 Typically, the two phases alternate in a Kalman filter, with the prediction
advancing the state until the next scheduled observation, and the
update incorporating the observation/measurements
Kalman Filter (how it actually works?)

 Summarized in one figure

Image source: Mahmoud Elsanhoury, University of Vaasa

Example of a Kalman filter
 Applied to two-dimensional location measurements

Source: Ali-Löytty, Collin, Sirola (2010). Mathematics for Positioning, TUT.

Navigation geometry (1)
 The transmitter-to-user geometry can have a large impact on the
accuracy of the position, velocity and time estimate
 A metric normally used for measuring this impact is DOP, which represents the
degree to which transmitter-user geometry dilutes the accuracy of the solution
 Information about the DOP is encoded in the geometry matrix G
 DOP-matrix H is denoted as

The DOPs are the link

between the pseudorange
 H11 − − −  errors and the position, velocity
 − H 22 − −  and time (PVT) estimation errors
H = (G T G ) −1 = 
 − − H 33 − 
 
 − − − H 44 
Navigation geometry (2)
 If operating in a local coordinate frame, the square root of the diagonal entries
of the H matrix are called the east DOP (EDOP), north DOP (NDOP), vertical
DOP (VDOP), and time DOP (TDOP).
 These DOPs are sometimes combined to form new DOPs such as the total geometry
DOP (GDOP), the three-dimensional position DOP (PDOP), and the two-dimensional
horizontal positioning DOP (HDOP)

EDOP = H11
GDOP = H11 + H 22 + H 33 + H 44
NDOP = H 22
PDOP = H11 + H 22 + H 33
VDOP = H 33
HDOP = H11 + H 22
TDOP = H 44
 In many land applications, when the receiver can be assumed to be on the
ground and its height constrained, the HDOP will improve
Navigation geometry (3)

Source:
P.H. Dana, Global Positioning
System Overview,
University of Texas, 2000.
Navigation geometry (4)

In GNSS, DOP is
the indicator of 19000 km

satellite geometry 20000 km

20500 miles

Accounts for
21000 km
location of each
satellite relative to
• A DOP of <2 is excellent
others • A DOP of 2-5 is good
• A DOP of 5-10 is moderate
Optimal accuracy • A DOP of 10-20 is fair
when DOP is LOW • A DOP >20 is poor

Image source:
http://en.wikipedia.org/wiki/Dilution_of_precision_(GPS)
Position Estimates with Noise and Bias

Source:
P.H. Dana, Global Positioning
System Overview,
University of Texas, 2000.
BIAS is for example multipath
BLUNDER (large bias) is for example echo-only signals or cross-correlation
Coordinate systems
Reference Frames (1)

 How do we give position information to objects?

(𝑥𝑟 , 𝑦𝑟 , 𝑧𝑟 )

Position is not an absolute value!

Reference Frames (2)

 We need a system to define a mathematical surface.

Two examples are CRS and TRS.

Terrestrial Reference System (TRS)

Celestial Reference System (CRS)

Reference Frames (3)

 A mathematical surface.
Reference Frames (4)

 A mathematical surface.

An example is the ellipsoid.

Reference Frames (5)

 A mathematical surface.

An example is the ellipsoid.

a = semi-major axis

b b = semi-minor axis

f = flattening = (a-b)/a

a 𝜔 = angular velocity
a
Reference Frames (6)

 …
Reference Frames (7)

 Cartesian Axis

Orign
Y
Scale

X
Reference Frames (8)

 Cartesian Axis

Z Orientation

Orign
Y

?
Scale

X
Reference Frames (9)

 Cartesian Axis
Orientation
in the Earth’s
Z (N) rotation axis

Orign
in the Earth’s Y (star)
center of mass

X Metric
Scale
Reference Frames (10)

 Cartesian Axis

It is the base to compute latitude,

longitude and altitude.
Various coordinate frames in navigation (1)

Source: M. Vermeer (2020), Methods of Navigation, Aalto University.

Various coordinate frames in navigation (2)

▸ Earth-centered (ECEF), local (ENU, NED), geodetic systems (WGS84)

▸ Transformation between systems is important for autonomous navigation

▸ mapping and localization in robotics
▸ georeferencing sensor data (LIDAR, cameras, radar)
Coordinate frames (1)

 The inertial frame is a reference frame that does not

accelerate or rotate with respect to the rest of the
universe (inertial frame i)
 In navigation applications, an Earth-centered inertial
(ECI) frame is used, which has its origin at Earth’s center
of mass but does not rotate or accelerate with Earth
 The Earth frame originates at the Earth’s center of
mass and has axes aligned to the Greenwich meridian
and the Earth’s mean spin axis, with the third axis
forming an orthogonal frame (Earth frame e)
Coordinate frames (2)

 The navigation frame is the locally defined frame

whose axes often point in the north, east, and down
directions (navigation frame n), often fixed with Earth
frame
 Aeroplanes
 Buildings
 The body frame is formed by the axes of the IMU,
which, in turn, are assumed to align roughly with the
front, right, and down directions of motions of the
vehicle being positioned (body frame b)
Coordinate frames (3)
 Changing coordinate frames:

Source:
J. Collin, Satelliittipaikannuksen
perusteet, Inertianavigointi-esitelmä,
TUT, 2006, 37 s.
Global and local coordinate systems – radar example

https://se.mathworks.com/help/phased/ug/global-and-local-coordinate-systems.html
Geocentric reference frames

• GPS produces coordinates in the WGS84 (World Geodetic System -84)

reference frame, the geocentric frame originally used by the GPS
• It is maintained by the US Department of Defense, and there have
been a number of versions

• States position and velocity in Cartesian coordinate system (XYZ)

• Earth-Centered Earth-Fixed
Coordinate System (ECEF)
• Origin at the center of mass of
the Earth
• XY-plane is coincident with
Earth’s equatorial plane

Source: T. Herring, Modern Navigation, MIT, 2009.

Coordinate Systems (cont.)

 Cartesian coordinates are usually transformed into

latitude, longitude and height (Curvilinear coordinates)
 For that transformation we need to have a physical
model describing the Earth
 World Geodetic System 1984 (WGS84)
 One part is detailed model of Earth’s gravitational
irregularities, needed also for ephemeris
 Ellipsoidal model of Earth’s shape
 INSPIRE directive defines the European coordinate system
(EUREF)
 Simple geocentric latitude and longitude
 The easiest form of latitude
and longitude to understand
is the spherical system:
 Latitude ϕ: angle between
the equatorial plane and the
point (P) R

 Longitude: angle between ϕ

the Greenwich meridian and
meridian of the location,
symbol λ λ
 Helsinki: Latitude 60.19, Longitude
24.94
 Vaasa: Latitude 63.09, Longitude
21.62
Geocentric relationship to XYZ
 One of the advantages of geocentric angles is that
the relationship to XYZ is easy
 R is taken to be radius of the sphere and H the
height above this radius
 However, this is only a simplification of the problem: Earth
is not a sphere, rather a geoid
Geoid

 Height is measured
relative to a surface
called geoid, which
corresponds to
global mean sea
level
Kaplan, E. and Hegarty, C. (2017)
 Topographic maps Understanding GPS/GNSS: Principles
and Applications. 3rd Edition, Artech
show heights relative House Publishers, London
to geoid, i.e. orthometric heights
 Geoid height N is the height of a point G above the
ellipsoid E. Orthometric height H is the height of point
P above the geoid G
Attitude angles of a vehicle

Source: M. Vermeer (2020), Methods of Navigation, Aalto University.

Time and timing systems
Time references (1)

▸ Fundamentals of timekeeping: Universal Coordinated

Time (UTC), Atomic Time (TAI), GPS Time

▸ Time synchronization in navigation (NTP, PTP)

▸ Role of atomic clocks in navigation

-> GNSS satellites provide timing accuracy
 Time references (2) Who defines TAI and Leap Seconds?
BIPM: International Bureau of Weights and Measures
Different time standards → IERS: International Earth Rotation and Reference Systems Service
http://leapsecond.com/java/gpsclock.htm

GPS = GPS Time

UTC = Universal Coordinated Time
TAI = Time Atomic International

TAI requires ~270 atomic clocks around the

world whose time is averaged.
GNSS timing (1)

▸ GNSS provide atomic Coordinated Universal Time (UTC) time to users

and enables precise synchronization for multiple applications

▸ In GPS, the current time is determined by the atomic clocks in the satellites
and modulated to as a navigation message on top of the coarse
acquisition (C/A) ranging code
▸ The receivers generate their own local replicas from the C/A codes received from each
satellite and estimate the time difference (delta) for aligning the local replicas to the
received copy
▸ The receivers also decode the navigation data for calculating the satellites position and
clock offsets and this information is used for estimating the 3D position and time

▸ The standard pulse-per-second (1-PPS) output of GNSS receivers are widely used in timing
and time synchronization due to the high accuracy and long-term stability
Source: J-P Kankaanpää (2021). GNSS Related Threats to Power Grid Applications, University of Vaasa.
GNSS timing (2)

▸ The Telecommunications sub-segment uses the GNSS timing

function for handover between base stations in wireless
communications, time slot management purposes and event
logging

▸ The Energy sub-segment, including power transmission, uses

GNSS timing in systems providing frequent measurements
relevant to the network status and to determine the location of
faults along a transmission line

▸ The Finance sub-segment uses GNSS to timestamp financial

transactions, allowing one to trace causal relationships and
synchronize financial computer systems
 GNSS timing (3)

Source: EUSPA (2020). GNSS User Technology Report.

Timing synchronization

▸ Synchronization challenges for autonomous

systems:
▸ Latency in sensor fusion
▸ Clock bias correction
▸ Decentralized networks
Some sources

• Kaplan, E. and Hegarty, C. (2017) Understanding GPS/GNSS: Principles

and Applications. 3rd Edition, Artech House Publishers, London

• M. Vermeer (2020). Methods of navigation – An introduction to

technological navigation. Aalto University publication series SCIENCE +
TECHNOLOGY 6/2020, https://users.aalto.fi/~mvermeer/nav-en.pdf

• Ali-Löytty, Collin, Sirola (2010). Mathematics for Positioning, TUT.

• M. S. Grewal, A. P. Andrews, Kalman Filtering, Theory and Practice Using

MATLAB, New York, Wiley, 2001

• J-P Kankaanpää (2021). GNSS Related Threats to Power Grid

Applications, Master’s thesis, University of Vaasa.

• EUSPA (2020). GNSS User Technology Report, Issue 3, 2020.

 Kiitos - Thanks!
[email protected]