Common Sensing Techniques for Reactive Robots

loutclankedAI and Robotics

Nov 13, 2013 (3 years and 10 months ago)

85 views

Common Sensing Techniques
for Reactive Robots

(12
-
11
-
7)

Sungmin Lee (
이성민
)

Division of Electronic Engineering, Chonbuk National University

Intelligent Systems & Robotics Lab.

http://robotics.jbnu.ac.kr

Chapter objectives


Describe difference between active and passive sensors



Describe the types of behavioral sensor fusion



Define each of the following terms in one or two sentences:
proprioception, exteroception, exproprioception,proximity sensor, logical
sensor, false positive, false negative, hue, saturation, computer vision



Describe the problems of specular reflection, cross talk, foreshortening,
and if given a 2D line drawing of surfaces, illustrate where each of these
problems would be likely to occur



If given a small interleaved RGB image and a range of color values for a
region, be able to 1) threshold on color and 2) construct a color
histogram


Contents

1.
Logical sensors


2.
Behavioral Sensor Fusion


3.
Attributes of a sensor


4.
Sensor Categories


5.
Computer vision


6.
Case study


7.
Summary




Motivation


Sensing is tightly coupled with acting in reactive systems,
so need to know about sensors



What sensors are out there?


Ultrasonics, cameras are traditional favorites


Sick laser ranger is gaining fast in popularity



How would you describe them (attributes)?



How would you decide which one to pick and use for an
application?

Logical Sensors


A unit of sensing or module (supplies a particular percept).


It consists of the signal processing and the software
processing.


Can be easily implemented as a perceptual schema.



Different sensors/perceptual schemas can produce the
same percept
-

motor schema doesn’t care!


Behavior can pick what’s available


Example: ring of IRs, ring of sonars


If sensor fails, then another can be substituted without
deliberation or explicit modeling


Conflicts in allocation can be solved by using logical
sensors (deliberation is required to assign)

Active vs. Passive (Example)



Active sensors


-

Sensor emits some form of
energy and then measures the
impact as a way of understanding
the environment


-

Ex.
Ultrasonics
, laser


Passive sensors


-

Sensor receives energy already in
the environment


-

Ex. Camera


Passive consume less energy,
but often signal
-
noise problems


Active often have restricted
environments

Stereo

Camera

pair

Thermal

sensor

Laser

ranger

Sonars

Bump

sensor

Behavioral Sensor Fusion

Sensor fusion
is a broad term used for any process that
combines information from multiple sensors into a single percept.

In some cases multiple sensors are used when a particular sensor
is too imprecise or noisy to give reliable data. Adding a second
sensor can give another “vote” for the percept.



When a sensor leads the robot to believe that a percept is
present, but it is not, the error is called a
false positive
.



The robot has made a positive identification of percept, but it was
false. Likewise, an error where the robot misses a percept is
known as a
false negative
.

False positive

False negative

Sensing Model

11

Sensor/Transducer

Behavior

Action

Sensing in Reactive Paradigm

Each behavior has its own dedicated sensing. One behavior
literally does not know what another behavior is doing or
perceiving.

Behavior

Behavior

Behavior

Perceptual

Schemas

M o t o r

Schemas

Behavioral Sensor Fusion:

-
sensor fission

This
sensor fission
in part as a take off on the connotations of
the word “fusion” in nuclear physics. In nuclear fusion, energy is
created by forcing atoms and particles together, while in fission,
energy is creating by separating atoms and particles.


Perceptual

S c h e ma

M o t o r

Schema

Behavioral Sensor Fusion:

-
action
-
oriented sensor fusion


This type of sensor fusion is called
action
-
oriented sensor fusion

to emphasize that the sensor data is being transformed into a
behavior
-
specific representation in order to support a particular
action, not for constructing a world model.


Perceptual

S c h e ma

M o t o r

Schema

Behavioral Sensor Fusion:

-
sensor fashion


Sensor fashion
, an alliterative name intended to imply the robot
was changing sensors with changing circumstances just as
people change styles of clothes with the seasons.

Designing a Sensor Suite

-
Attributes of a sensor


Field of view, range
: does it cover the “right” area


Accuracy & repeatability
: how well does it work?


Responsiveness in target domain
: how well does it work for
this domain?


Power consumption
: may suck the batteries dry too fast


Reliability
: can be a bit flakey, vulnerable


Size
: always a concern!


Computational Complexity
: can you process it fast enough?


Interpretation Reliability
: do you believe what it’s saying?

Should be considered for the entire sensing suite :


Simplicity


Modularity


Redundancy

-

physical redundancy


(
there are several instances of physically identical sensors on the
robot.)

-
logical redundancy


(
another sensor using a different sensing modality can
produce the same percept or releaser.)

-

fault tolerance


Designing a Sensor Suite

-
Attributes of a sensor suite

Sensor Categories


Proprioceptive


Inertial Navigation System(INS)


Global Positioning System(GPS)


Exteroceptive


Proximity


Range


Contact


Computer Vision


Proprioceptive Sensors(1)

-
Inertial navigation system (INS)


Measure movements electronically through miniature
accelerometers INS can provide accurate dead reckoning to 0.1
percent of the distance traveled.

However, this technology is unsuitable for mobile robots for
several reasons.(cost, Size, etc)

MQ
-
9 Reaper

GPS systems work by receiving signals from satellites orbiting the
Earth.

GPS is not complete solutions to the dead reckoning problem in
mobile robots.

GPS does not work indoors(environmental limit)

Proprioceptive Sensors(2)

-
Global Positioning System (GPS)



Proximity Sensors(1)

-
Sonar or ultrasonic



Sonar refers to any system for using sound to
measure range. (use a sonar for underwater
vehicles ).



Ground vehicles commonly use sonar with an
ultrasonic frequency.



Ultrasonic sensors generate high frequency
sound waves and evaluate the echo which is
received back by the sensor. Sensors calculate
the time interval between sending the signal
and receiving the echo to determine the
distance to an object.



Ultrasonic is possibly the most common
sensor on commercial robots operating


Polaroid
ultrasonic
transducer

chairs, tables


legs, edges too thin for resolution

Proximity Sensors(1)

-

Three problems with sonar range readings


foreshortenin
g

cross
-
talk

specular reflection

Maps produced by a mobile robot using
sonars

in: a.) a lab and b.) a
hallway. (The black line is the path of the robot.)

Proximity Sensors(1)

-

sonar maps


lab

hallway


Power consumption


High


Reliability


Lots of problems


Size


Size of a Half dollar, board is similar size and can be creatively
packaged


Computational Complexity


Low; doesn’t give much information


Interpretation Reliability


poor


Proximity Sensors(1)

-

Attributes of ultrasonic



Physics

: active sensor, works on time of flight


Advantages
: range, inexpensive ($30 US), small


Disadvantages
: specular reflection, crosstalk, foreshortening,
high power consumption, low resolution



Proximity Sensors(1)

-

Ultrasonic Summary


They emit near
-
infrared energy and measure whether any
significant amount of the IR light is returned.

These often fail in practice because the light emitted is often
“washed out” by bright ambient lighting or is absorbed by dark
materials (i.e., the environment has too much noise).

Proximity Sensors(2)

-

Infrared ray (IR)


Sharp GP2Y0A21YK

Popular class of robotic sensing is tactile, or touch, done with
bump and feeler sensors.

The sensitivity of a bump sensor can be adjusted for different
contact pressures


Proximity Sensors(3)

-

Bump and feeler sensors


Roomba

500

Bump

Computer Vision

-

Definition

Computer vision refers to processing data from any modality
which uses the electromagnetic spectrum which produces an
image.


face
recognition

Computer Vision

-

Attributes


Physics

: light reflecting off of surfaces, respond to wavelenght



Field of view, range
: depends on lens; lens typical have a
different VFOV and HFOV (vertical, horizontal)



Accuracy & repeatability
: good



Responsiveness in target domain
: depends on lighting source,
inherent constrast between objects of interest



Power consumption
: low



Reliability
: good



Size
: can be miniaturized



Interpretation Reliability
: good



A charge
-
coupled device (CCD) is a
device for the movement of electrical
charge, usually from within the device to
an area where the charge can be
manipulated, for example conversion into
a digital value.

Computer Vision

-

CCD cameras



CCD sensors typically produce less NOISE.



CCD sensors typically are more light
-
sensitive.



CMOS sensors use far less power.



CMOS sensor cost less to produce.


RGB (red, green, blue) is the NTSC output


Poor color constancy in “real world”


H,S,I (hue, saturation, intensity) has theoretical color
constancy


But not with conversion from RGB to HSI


Alternatives SCT (Spherical Coordinate Transform)


That color space was designed to transform RGB data to a
color space that more closely duplicates the response of
the human eye.

Computer Vision

-

Color planes

Original image

RGB

HSI

SCT


For reactive applications:


Color segmentation


Imprint on a color region, then follow it


(or remember it)


Color
histogramming


Imprint on a region with a distribution of color,
then follow it (or remember it)

Computer Vision

-

Common Vision Algorithms

Range from Vision

-
Stereo camera pairs



Using two cameras to extract
range data is often referred to as
range from stereo, stereo
disparity, binocular vision, or just
plain “stereo.” One way to extract
depth is to try to superimpose a
camera over each eye.




Each camera finds the same
point in each image, turns itself
to center that point in the
image, then measures the
relative angle. The cameras are
known as the stereo pair.

Ways of extracting depth from a pair of cameras

Light striping, light stripers or structured light detectors work by
projecting a colored line (or stripe), grid, or pattern of dots on the
environment. Then a regular vision camera observes how the pattern is
distorted in the image.

Range from Vision

-
Light stripers

Range from Vision

-
Laser ranging(Sick)


Accuracy & repeatability
-

Excellent results


Responsiveness in
target domain


Power consumption



-

High; reduce battery run time by half


Reliability

-

good


Size

-

A bit large


Computational Complexity


Not bad until try to “stack up”


Interpretation Reliability


Much better than any other ranger

flat surface

an obstacle

a negative obstacle

SICK PLS100


180
o

plane


Advantages: high accuracy, coverage


Disadvantages: 2D, resistant to miniaturization, cost
($13,000 US)


Range from Vision

-
Laser Ranger Summary

NASA/CMU Nomad robot

(Carnegie Mellon University )

Case Study :

Hors d’Oeuvres, Anyone?

(Borg Shark and Puffer Fish)


Camera pair (redundant):

Face color

Laser range:

Count treat

removal

Sonars
:

Avoid obstacles,

If blocked,

Puffed up


Digital thermometer:

“Face” temperature check

Sensor fusion:

Reduced

False positives,

False negatives

From 27.5% to 0%


State diagrams for the Borg Shark

sonar

shaft

encoders

map:

evidence grid

waypoint

navigation:

move to goal,

avoid

serving food:

finding faces,

counting treat

removal

thermal

vision

laser range

sonar

awaiting

refill:

finding faces

vision

thermal

at waypoint

OR

food removed

time limit

serving exceeded

food

depleted

full

tray

Summary


Design of a sensor suite requires careful consideration


Almost all robots will have proprioception, but exteroception
needs to be closely matched to the task and the environment


Most common exteroceptive sensors on mobile robots are:


Ultrasonics


Computer vision


Laser range


Color vision can be hard, almost all vision is computationally
expensive unless focus on affordances


Borg shark and Puffer fish with color plus heat


Polly and texture



Thank you