Brainscape's Knowledge GenomeTM

Browse over 1 million classes created by top students, professors, publishers, and experts.


See full index

Human Robot Interaction > Lecture 2 - PR1 > Flashcards

Lecture 2 - PR1 Flashcards

1
Q

Robots must deal with many uncertainties, name 4:

A

Noisy sensors
Unknown location
Outdated maps
Inaccurate odometry and dead reckoning

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

An event for which the outcome is uncertain is represented using a ….

A

random variable

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What is the difference between discrete and continuous random variables

A

The outcomes are either discrete values (so 1,2,3,4,5,6 in case of the dice), and for continuous random variables the outcomes are continuous values

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What is the likelihood in Bayes’ function?

A

It reflects sensory information

Function of hypothesis, will typically not integrate to 1

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

What is the Prior?

A

Independent of observations

Reflects prior knowledge about the hypothesis

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

What is the Posterior?

A

Reflects the belief in the hypothesis

Takes prior knowledge into account

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

What is P(open|z)

A

is diagnostic

what is the probability that the door is open, given the measurement?
This is the posterior belief

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

What is p(z|open)

A

Is causal
Is the likelihood
Probability of a measurement, given the state of the world (that the door is open)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

What is the Markov assumption

A

The Markov assumption, is an assumption made in Bayesian probability theory, that every new measurement, is conditionally independent of its previous measurement.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

What 3 things underly the Markov Assumption?

A
  • there is a static world
  • we are independent of noise
  • perfect model with no estimation errors
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

Explain why Bayes’ filter is a recursive loop

A

Because you continuously delete all but the previous measurement. When finally you get back to an equation with your last belief in it. And that is your new belief. This constantly keeps updating

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

Why do domestic environments pose additional problems?

A

Cluttered
Dynamic
People

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What is the difference between local and global robot navigation

A

Global:
Map-based
Deliberative
Slow

Local

  • Sensory-based
  • Reactive
  • Fast
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

Difference between model based and behavior based navigation?

A

Model based:

  • complete modelling
  • function based
  • serial process

Behavior based:

  • sparse or no modeling
  • behavior based
  • bottum up
  • parallel processing
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Methods for navigation:

  1. Incrementally
  2. Modifying the environments

What are the limitations of these methods?

A
  1. dead reckoning -> you go into the environment without any information. Therefore the errors you make become bigger and bigger
  2. You need to place inductive or optical tracks, or use reflectors or other things. However, this is expensive and very inflexible
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

In behavior-based control architecture the robot’s actions are determined by …..

A

a set of parallel, reactive behaviors which map sensory input and state to actions

17
Q

What is subsumption architecture?

A

One of the earliest behavior-based navigation architectures

Behaviors are arranged in a strict priority order

18
Q

What is the Kalman filter and what does it assume?

A

It is a particular implementation of a Bayesian filter with a number of simplifying assumptions

Assumes:

  • all random variables are normally distributed
  • the state updates are linear functions

Consequences:
- the dependent variable is also normally distributed

19
Q

What is capital K in the Kalman filter?

A

The Kalman gain, corrects for errors

20
Q

Difference between Kalman and the Particle Filter

A

Particle filter does not update parameters of exact distributions.
You can easily check multiple hypotheses
Each particle can be thought of as a hypothesis about the true state of the world

21
Q

How does localization work for particle filters?

A
  1. particles are propagated according to the motion model
  2. then they are weighted according to the likelihood of observations
  3. in re-sampling, new particles are drawn with a probability proportional to the likelihood of that observation
22
Q

Person detection and localization happens via 5 visual cues:

A
  1. motion
  2. recent color
  3. static shape
  4. recent shape
  5. TLD algorithm

Human Robot Interaction (7 decks)

Brainscape helps you reach your goals faster, through stronger study habits.
© 2024 Bold Learning Solutions. Terms and Conditions