Sensors

Question 1

Perception

Answer 1

Can be either raw data, features, objects or places/situations depending on the processing of the raw data becoming from sensors.

Question 2

Difference between sensing and sensors in AI

Answer 2

Sensors provide raw data, while sensing is the combination of algorithms and sensors that produce a percept or world model.

Question 3

What is a sensor?

Answer 3

Is a device that measures some attributes of the world. Sensors use transducers to change the input signal into an analog or digital form capable of being used by a robot.

Question 4

What is a transducer?

Answer 4

Is a mechanical part of a sensor that transforms the energy associated to what is being measured into another form of energy, that is the sensor raw data used by the robot system algorithms.

Question 5

What is the perceptual schema?

Answer 5

Is one of the 5 subsystems. Its goal is to generates percepts using sensors raw data.

Question 6

Active sensor vs passive sensor

Answer 6

• active sensor: put energy on the environment either to change the energy or enhance it (e.g. camera with flash, ultrasonic sensor, laser rangefinder)
• passive sensor: rely on the environment to provide the medium for observation (e.g. camera without flash)

Question 7

Active sensing

Answer 7

Connotes a system using an effector to dynamically position a sensor to improve data gathering.

Question 8

Sensor modality

Answer 8

Sensors which measure the same form of energy and process it in similar ways form a sensor modality. “Modality” refers to the raw input used by the sensors.

Question 9

The 3 types of perception

Answer 9

• proprioception: locate the position of limbs and joints of the robot or to determine how much they have moved (essential for control of the robot platform)
• exteroception: detect objects in the world and often the distance to those objects (essential to enable the robot to move and act in the world)
• exproprioception: detect the position of the robot body or parts relative to the layout of the environment (important for manipulation and for reasoning about why the robot is not moving)

Question 10

Heading sensors and dead reckoning

Answer 10

Heading sensors can be proprioceptive or exteroceptive. They are used to determine the robot’s orientation and inclination. They allow us, together with appropriate velocity information, to integrate the movement to a position estimate. This procedure, which has its roots in vessel and ship navigation, is called dead reckoning.

Heading sensors:
* compasses: they measures the direction of the magnetic field (exteroceptive)
* inclinometers: used to measure the angle of slop (proprioceptive)
* gyroscopes: used to measure the orientation (proprioceptive)
* accelerometers: measure all external forces acting upon it (proprioceptive)

Question 11

Shaft encoders, INS and GPS

Answer 11

• shaft (or optical) encoders: similar to odometer in cars, they compute the number of turns of the robot’s wheel knowing the number of turns of the motor and the size of the wheels
• INS (inertial navigation system): uses gyroscopes and accelerometers to provide dead reckoning. Hard bumps and sudden turns can introduce errors. An INS estimates the 6 DOF pose of the vehicle.
• GPS (global positioning system) receivers: common in robots that work outdoors. Passive but exteroceptive sensors.

Question 12

Active ranging sensors

Answer 12

Are sensors that use an active sensing to measure the distance or range to an object or target.

Most popular sensors in mobile robotics.

Associated with time-of-flight active range sensors which make use of the propagaton speed of sound or an electromagnetic wave to compute the distances.

d = c*t

Question 13

Proximity sensors

Answer 13

Directly measure the relative distance (range) between the sensor and objects in the env. Proximity sensors generally connote measuring short distances, on the order of 1 meter.

Range sensors (e.g. laser range finder) are used for longer distances.

Tactile sensors are also used as proximity sensors.

Question 14

Infrared sensors

Answer 14

Are the most popular type of proximity sensors.

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

The common IR sensors often fail because the light emitted is often “washed out” by bright ambient lightining or is absorbed by dark materials.

Question 15

Triangulation active ranging sensors

Answer 15

These sensors project a known light pattern (structured light) onto the env and through the reflection of this pattern (combined with geometric values) the system can use simple triangulation to establish range measurements (e.g. Kinect sensor).

Question 16

Examples of proprioceptive & exteroceptive sensors

Answer 16

Proprioceptive sensors:
* gyroscopes
* inclinometers
* accelerometers
* inertial navigation system (INS): combination of the two above
* shaft (or optical) encoders
* force/torque sensors
* tactile sensors

Exteroceptive sensors:
* compasses
* proximity sensors (IR)
* range sensors (laser range finder aka lidar, ultrasonic)
* triangulation active ranging (structured light)

Question 17

Computer vision

Answer 17

Is the primary source of general exteroceptive sensing for direct perception and object recognition.

Refers to processing data from any modality which uses the electromagnetic spectrum to produce an image.

Question 18

HUE

Answer 18

Is the dominant wavelength. It is used because does not change with the robot’s relative position or the object’s shape.

Cameras that are sensitive to the hue works on the HSV (hue, saturation, value/intensity) representation of the color.

Challenging to implement these cameras on robotics because:
* highly expensive
* software is complex and leads to singularities

Question 19

Logical sensors

Answer 19

Is a unit of sensing, or module, that supplies a particular percept. It is the functional building block for perception.

It consists of:
* the signal processing from the physical sensors
* the software processing needed to extract the percept

A logical sensor can be easily implemented as a perceptual schema.

Question 20

Logical sensor

Answer 20

A logical sensor is composed by a set of logically equivalent sensors that can return the same percept data structure.

Question 21

Three combination of sensors

Answer 21

• redundant: reduce noise and failures
• coordinate: multiple sensors that are calibrated and synchronized to work together (more complex measurements)
• complementary: use of multiple sensors that are of different types (more robust understanding of the environment)

Question 22

Action-oriented sensor fusion, fashion and fission

Answer 22

• action-oriented sensor fusion: is a broad term used for any process that combines information from multiple sensors into a single percept. It can be used for creating high-level features or making less noisy information
• fashion: change sensors with changing circumstances. Only one percept is taken for generating a behavior.
• fission: the best generated behavior is taken.

draw the diagrams

Question 23

False positive vs false negative

Answer 23

• false positive: the robot senses a percept but its not present
• false negative: the robot doesn’t sense a present percept

Question 24

Sensor suite and attributes for constructing it

Answer 24

Is a set of sensors for a particular robot.

In order to construct a sensor suite, some attributes should be considered for each sensor:

• FOV
• accuracy, repeatability, stability, resolution (sensitive)
• responsiveness in the target domain: the env must allow the signal of interest to be extracted from noise and interference
• power consumption
• reliability: physical limitations
• size
• computational complexity

Question 25

Basic sensor response ratings

Answer 25

Are some measurements that help to compute the performance of a sensor.

• dynamic range: ration between the maximum measurable input value and the lowest (in decibels)
• resolution: minimum difference between two values that can be detected by a sensor
• linearity: f(ax + by) = af(x) + bf(y) for any a and b (x and y are the input signals)
• bandwidth or frequency: speed at which a sensor can provide raw data

Question 26

Accuracy and precision (formulas)

Answer 26

• accuracy = 1 - |error|/v where v is the true value
• precision = range/o where range is the output’s range and o is the standard dev.

Question 27

Types of errors in sensors measurements

Answer 27

• linearity error: maximum deviation of the measured output from the straight line that best fits the real characteristics
• offset error: value of the measured output for zero input
• resolution error: maximum variation of the input quantity producing no variation of the measured output

Question 28

What sensors are essential for a robot?

Answer 28

In general, robots will need both proprioceptive sensors to keep track of the pose of its physical body and exteroceptive sensors to navigate and recognize objects.