Field and Service Robotics

Odometry sensors

Navigation (internal) Sensors

To sense robots own state
z Magnetic compass (absolute heading)
z Gyro (angular speed => change of heading)
z Acceleration sensors (acceleration)
z tako, encoder (speed, distance)
z syncro, resolver (speed, position)

Dead-reckoning
z
z
z
z
z
z
z
z

Maritime term from deduced reckoning

Murtoviivasuunnistus in Finnish
The position is calculated on the basis of previous
position, heading and travelled distance
No external beacons needed
Very sensitive to the heading measurement!
Error is accumulating all the time!!
With basic sensors a very low cost alternative
do always!
High accuracy costs a lot!

Dead-reckoning
z

The distance is measured with odometry

easy with wheels and tracks
difficult with legs and in maritime and airborne applications

The heading can be defined by using:

Measurements and vehicle kinematics
z steering angle + odometry measurement (ackermann)
z Odometry (skid steering)
Direct measurement
z compass
z gyro
z etc.

Note: different angles!

Alpha = steering angle
Theta = body angle

Kinematics

D=

Dl + Dr
2

Dl Dr
d

= the total distance

= heading

cot sa =

d
+ cot i
2l

cot i cot 0 =

d
l

Here:
theta = stering angle

Ackerman Steering

Differential Steering (J2B2)

Note: Dynamics (mass inertia) affects
always, especially when speed increases!

No Kinematics
z
z

Tracked and wheeled

skid-steered vehicles
Wheels have to slide
sideways, while
turning
Surface properties and
CoG say the last
word during the
steering process

Error accumulation
z

Distance error from

slippery etc.
incrementally moves
the calculated final
position
Heading error from
steering, gyro, etc. has
much bigger effect
heading is critical!!

Odometers
z
z
z
z
z
z
z
z

tachometer (DC
motor)
potentiometer
synchro
resolver
optical encoder
magnetic encoder
inductive encoder
kapasitive encoder

The most used

odometers are:
z optical encoders
z resolvers
z potentiometers

Potentiometer
z
z
z

low-cost, easy to use

resistace/voltage proportional to the absolute angle/position
not for continuous rotation, mechanical wearing, nonlinear in accurate
measurements

DC-Tachometer
z

Like normal DC-motor (DC with brushes,

AC without)
z Output voltage proportional to the rotation
speed
z cheap, output also in low speeds
z wearing brushes (not in brushless), analog
output

Synchro
z
z

Electro mechanical system which transfers the

angle information with high accuracy
the receiver rotor will follow the transmitter rotor
when same alternating voltage is supplied to both
rotors

Resolver
z
z

Special configuration of synchro

The output phase is proportional to shaft absolute
angle and output (amplitude)/frequency is
proportional to the shaft speed (ideal feedback
sensor for brushless DC-motors)

Variable reluctance resolver

z
z

z
z
z
z

Rotor consists of only electromagnetic steel sheets

One-phase of exciting coil and
two-phase output coils are
wound on the stator core
The contour of rotor is made as
ellipsoid i.e. air-gap between
the stator and rotor is varied in
sinusoid depending on the shaft
angle
No brushes or sliprings
Extremely robust
Both speed and absolute angle
output
Used in Toyota Prius
http://www.scribd.com/doc/38612704/Tamagawa

Resolver
z
z
z
z
z
z

Reliable
Accurate
Dual output
Analog
high price
(brushes wear)

Brushless resolvers are

also available
z bigger
z no mechanical wearing
z bigger power consuption
Variable reluctance
resolvers are the future

Optical Encoders
z

accurate
z enduring
z low-price (incremental)
z usually not as robust as resolvers

Incremental Optical Encoder

z
z
z

Primary quantity is relative distance/angle (pulse)

The most simple is one channel takometer
Adding an other channel also the direction of rotation can
be sensed

Absolute Encoder
z

Measures the absolute position/angle

z More expensive than incremental one
z Best for slow rotations with high accuracy

Absolute Encoder
The most used configurations are:
z A: Gray-code
z B: Binary code

Ground Speed Radar

z
z

The wheels and tracks always slide => odemetry is not

accurate
Doppler radar gives the real groung speed, proportional to
the speed difference between the radar and the target

Sensor Fusion
z

Usually there is several (unaccurate)

positioning measurements available like
odometry, steering angle, gyro, (D)GPS,
acceleration sensors etc.
z To get the best possible estimate of the
position the sensor data has to be fused
z Kalman-filter is the most used estimation
algorithm in robotics