CS 7638 Exam Questions with Verified Correct
Answers
explain localization (histogram filter)
Localization assumes that the map of the environment is already known. You then use
measurements and movements to build a belief of where the robot is in the environment.
What is the limit distribution of a robot? (infinite movements but no measurements)
The result will be a uniform distribution. So the more we more, the more information we will
lose about the environment and our location.
What is the formula for entropy?
-SUM(P(Xi) log(P(Xi))
In Histogram Filter, how is the Sense/measurement function calculated?
Using Products (multiplication) followed by normalization. (Using baye's theorem)
In Histogram Filter,, how is the Move/motion function calculated?
Using convolution (addition). (using total probability). The operation of a weighted sum is
called a convolution.
How does the memory scale wrt. the number of state variables in a Localization
method? (linearly? Quadratically? exponentially)
Exponentially. If each variable had 20 possible values (bins) then the joint probability table
will be 20^n where n is the number state dimensions.
What is the number of one issue with localization grid (histogram) method?
The memory allocation is exponential.
, What is standford's first self driving car called?
Junior - uses lasers and radar
What is Kalman filter used for?
Tracking & localization of the robot. It helps track the movement of other objects in our
environment so we don't collide with them.
Is Kalman Filter discrete or continious?
continious
Is Kalman filter uni-modal or multi-modal?
uni-modal (using Gaussians)
Is Monte Carlo Localization discrete or continuous?
discrete
Is Monte Carlo Localization Uni-modal or Multi-modal? In (Histogram Filter)
Multi-modal
Are particle filters discrete or continious?
Continious
Are particle filters uni-modal or multi-modal?
multi-modal
Explain a gaussian
A gaussian is a normal distribution (uni-modal) where the area underneath sums up to 1.
In Kalman filters, how is measurement update calculated?
Answers
explain localization (histogram filter)
Localization assumes that the map of the environment is already known. You then use
measurements and movements to build a belief of where the robot is in the environment.
What is the limit distribution of a robot? (infinite movements but no measurements)
The result will be a uniform distribution. So the more we more, the more information we will
lose about the environment and our location.
What is the formula for entropy?
-SUM(P(Xi) log(P(Xi))
In Histogram Filter, how is the Sense/measurement function calculated?
Using Products (multiplication) followed by normalization. (Using baye's theorem)
In Histogram Filter,, how is the Move/motion function calculated?
Using convolution (addition). (using total probability). The operation of a weighted sum is
called a convolution.
How does the memory scale wrt. the number of state variables in a Localization
method? (linearly? Quadratically? exponentially)
Exponentially. If each variable had 20 possible values (bins) then the joint probability table
will be 20^n where n is the number state dimensions.
What is the number of one issue with localization grid (histogram) method?
The memory allocation is exponential.
, What is standford's first self driving car called?
Junior - uses lasers and radar
What is Kalman filter used for?
Tracking & localization of the robot. It helps track the movement of other objects in our
environment so we don't collide with them.
Is Kalman Filter discrete or continious?
continious
Is Kalman filter uni-modal or multi-modal?
uni-modal (using Gaussians)
Is Monte Carlo Localization discrete or continuous?
discrete
Is Monte Carlo Localization Uni-modal or Multi-modal? In (Histogram Filter)
Multi-modal
Are particle filters discrete or continious?
Continious
Are particle filters uni-modal or multi-modal?
multi-modal
Explain a gaussian
A gaussian is a normal distribution (uni-modal) where the area underneath sums up to 1.
In Kalman filters, how is measurement update calculated?
