Lect 21-22

Question 1

How to link the actions and perceptions?

Answer 1

With using software architectures, we can link the perceptions and actions

Question 2

What is perception?

Answer 2

The robot perceives the world and sends it to the actions and architectures.

Question 3

Why is perception is hard?

Answer 3

Because it is very hard to implement it. From the pixels the robot wont be able to understand what is going on but for a human it is easy to look at a picture and see if there is a danger. We need a common sense. In order to do it we need to extract meaningfully features.

Question 4

Why sensors are important?

Answer 4

Sensors are the key technology for perceiving the environment.
We need to understand the physical principles in order to:
-Choose the proper sensor
-Properly model the sensor system.
- Define the needs in collaboration with sensor system.
We need to ENGINEER THE ENVIRONMENT

Question 5

What are the challenges for the robots?

Answer 5

Seeing feeling undertsanding, partially available environment act properly.
In partially available environments humans are really good in reconstructing what is missing. But for a robot it is a very difficult task.
Occlusions: one object dissapearing behind another object.

Question 6

What is the structure of perception, what is the structure of sensor processing?

Answer 6

Places and situations
|
Objects
|
Feature Extraction
|
Raw data

When going from bottom to up, compressing the input data is higher. And more abstract representation on the input data.

Question 7

What is the structure of perception?

Answer 7

First from the sensors, the raw data is coming, we need to apply feature extractionAfter that we can apply segmentation which is segmenting the features from the rest of the data to recognize and characterize the objects. With further processing, we can find the places or situations.

Question 8

What is sensor, sensing and sensor fusion?

Answer 8

Sensor: Sensors provide raw data
Sensing: the combinations of algorithms and sensor data to get percept
Sensor fusion: Sensing mechanism which allows us using multiple sensors in order to have percepts.

Question 9

What are the common sensors used in mobile robotics?

Answer 9

Tactile/Bumper Sensors:
Detection of physical contact
GPS:
Global localization and navigation
Wheel encoders:
Local motion estimation (odometry)
Inertial Measurement Unit:
The orientation and acceleration of the robot
Laser Scans:
Avoid obstacles, motion estimation
Cameras: 
Texture information,motion estimation

Question 10

Categorization of sensors:

Answer 10

The categorization of sensors depend on the input they provide.

1) Proprioceptive: Sensing stimuli within an organism
2) Exteroceptive: Sensing stimuli external to the organism.
3) Exproprioceptive: Sense the position of external object, relative to the parts of the robot.

With exteroceptive we can identify the obstacles but just if our camera is seeing that obstacles. We can think of exproprioceptive as an external camera which is filming the whole environment

4) Input
4. 1 : Active: emmiting energy from the environment and measuring the reaction
4. 2 Passieve: Energy coming from the environment
5) Output
5. 1 Image non image
5. 2 Matrix non matrix

Camera is a passive sensor under the sun light because it is collecting energy from the environment, but when the sun is not there it becomes an active sensor.

Question 11

Examples of proprioceptive sensors?

Answer 11

All kind of encoders are proprioceptive sensors. 
Gyroscope (P) Heading sensor
Brush encoders(P)
Magnetic encoders(A)
Inductive encoders(A)
Optical encoders(A)

Question 12

Examples of exteroceptive sensors?

Answer 12

Bumping sensors(P)
CCD CMOS cameras (P)
Optical barriers(A)
GPS (A)
Inclinometer (A/P)
Laser range finder(a)
ultrasonic sensor(A)

Question 13

Characterizing sensor performance, what is dynamic range,resolution, range, bandwith/frequency, linearity?

Answer 13

Dynamic range finder: ratio between the upper and lower limit of a sensor
Range
resolution: lower limit of dynamic range finder
Linearity: output signal as a function of input signal
Bandwith/Frequency: How fast your sensor can provide readings and how fast it can update.

Question 14

What is perceptive horizon?

Answer 14

What our robot can sense. It depends on the occlusion and range

Question 15

What is sensitivity?

Answer 15

How much should the input change to get a different output

Question 16

What is cross sensitivity(cross talk)?

Answer 16

1) Sonar: you emmit an ultrasound and you listen for the echo coming back, but there are other sonars emmiting ultrasound, therefore your reading is influenced with the other sonars.
2) Compas: measures the magnetic field,