CS 7638 Final Exam Questions and Answers Graded A+
What is localization for? - Answers To determine a robot's location within a certain level of
accuracy
What happens to the new mean when you have two equal standard deviation's and you combine
them? - Answers The new mean is located in the middle
What happens to the new mean when you have two unequal standard deviation's and you
combine them? - Answers The new mean goes towards the smaller variance
A valid probability distribution must sum to: - Answers 100%
Is the sense function calculated used Bayes Rule or Law of Total Probability? - Answers Bayes
Rule (Normalized Product)
Is the move function calculated used Bayes Rule or Law of Total Probability? - Answers Law of
Total Probability (Convolution)
Is measurement a product or a convolution? - Answers Product (Bayes Rule)
Is motion a product or a convolution? - Answers Convolution (Law of Total Probability)
How do you pick the Gaussian distribution with the largest sigma? - Answers A wider
distribution has a larger sigma, while a skinny distribution has the smallest sigma
Are Histogram Filters discrete or continuous? - Answers Discrete
Can a Histogram Filter be multimodal, or just unimodal? - Answers Can be multimodal
When applied to robots, do you believe the Histogram Filter is exact or approximate? - Answers
Approximate
In regards to Histogram Filters, when it comes to scaling in the number of dimensions of the
state space, which of the following is the amount of storage that must be assigned? - Answers
Exponential
What happens to the probabilities of movements in a cyclic world as the number of movements
approach infinity? - Answers They become equally distributed
Are Kalman Filters discrete or continuous? - Answers Continuous
Can a Kalman Filter be multimodal, or just unimodal? - Answers Cannot be multimodal (Just
unimodal)
In regards to Kalman Filters, when it comes to scaling in the number of dimensions of the state
space, which of the following is the amount of storage that must be assigned? - Answers
Quadratic
, When applied to robots, do you believe the Kalman Filter is exact or approximate? - Answers
Approximate
When would a narrow Gaussian distribution (low variance) in the state (x matrix) NOT be good
for the performance of a Kalman filter? - Answers When the mean of the prediction is far away
from the mean of the measurement
How do we calculate the next position of a robot? - Answers X = F * Xn-1
The Kalman gain is the weight applied to the measurements and the ___ estimate when
updating the state. - Answers Predicted state / current state
Is the Kalman gain calculated in the prediction step of the Kalman filter? - Answers False
Prediction or Measurement Step:
x' = F x + u - Answers Prediction
Prediction or Measurement Step:
P' = F P F^T - Answers Prediction
Prediction or Measurement Step:
y = z - H x - Answers Measurement
Prediction or Measurement Step:
S = H P H^T + R - Answers Measurement
Prediction or Measurement Step:
K = P H^T S^-1 - Answers Measurement
Prediction or Measurement Step:
x' = x + (K y) - Answers Measurement
Prediction or Measurement Step:
P' = (I - K H) P - Answers Measurement
What is the strategy for resampling in Particle Filters? - Answers Resample AFTER normalizing
our importance weights to determine the probability of resampling a particle
When applied to robots, do you believe the Particle Filter is exact or approximate? - Answers
Approximate
Can a Particle Filter be multimodal, or just unimodal? - Answers Can be multimodal
What is localization for? - Answers To determine a robot's location within a certain level of
accuracy
What happens to the new mean when you have two equal standard deviation's and you combine
them? - Answers The new mean is located in the middle
What happens to the new mean when you have two unequal standard deviation's and you
combine them? - Answers The new mean goes towards the smaller variance
A valid probability distribution must sum to: - Answers 100%
Is the sense function calculated used Bayes Rule or Law of Total Probability? - Answers Bayes
Rule (Normalized Product)
Is the move function calculated used Bayes Rule or Law of Total Probability? - Answers Law of
Total Probability (Convolution)
Is measurement a product or a convolution? - Answers Product (Bayes Rule)
Is motion a product or a convolution? - Answers Convolution (Law of Total Probability)
How do you pick the Gaussian distribution with the largest sigma? - Answers A wider
distribution has a larger sigma, while a skinny distribution has the smallest sigma
Are Histogram Filters discrete or continuous? - Answers Discrete
Can a Histogram Filter be multimodal, or just unimodal? - Answers Can be multimodal
When applied to robots, do you believe the Histogram Filter is exact or approximate? - Answers
Approximate
In regards to Histogram Filters, when it comes to scaling in the number of dimensions of the
state space, which of the following is the amount of storage that must be assigned? - Answers
Exponential
What happens to the probabilities of movements in a cyclic world as the number of movements
approach infinity? - Answers They become equally distributed
Are Kalman Filters discrete or continuous? - Answers Continuous
Can a Kalman Filter be multimodal, or just unimodal? - Answers Cannot be multimodal (Just
unimodal)
In regards to Kalman Filters, when it comes to scaling in the number of dimensions of the state
space, which of the following is the amount of storage that must be assigned? - Answers
Quadratic
, When applied to robots, do you believe the Kalman Filter is exact or approximate? - Answers
Approximate
When would a narrow Gaussian distribution (low variance) in the state (x matrix) NOT be good
for the performance of a Kalman filter? - Answers When the mean of the prediction is far away
from the mean of the measurement
How do we calculate the next position of a robot? - Answers X = F * Xn-1
The Kalman gain is the weight applied to the measurements and the ___ estimate when
updating the state. - Answers Predicted state / current state
Is the Kalman gain calculated in the prediction step of the Kalman filter? - Answers False
Prediction or Measurement Step:
x' = F x + u - Answers Prediction
Prediction or Measurement Step:
P' = F P F^T - Answers Prediction
Prediction or Measurement Step:
y = z - H x - Answers Measurement
Prediction or Measurement Step:
S = H P H^T + R - Answers Measurement
Prediction or Measurement Step:
K = P H^T S^-1 - Answers Measurement
Prediction or Measurement Step:
x' = x + (K y) - Answers Measurement
Prediction or Measurement Step:
P' = (I - K H) P - Answers Measurement
What is the strategy for resampling in Particle Filters? - Answers Resample AFTER normalizing
our importance weights to determine the probability of resampling a particle
When applied to robots, do you believe the Particle Filter is exact or approximate? - Answers
Approximate
Can a Particle Filter be multimodal, or just unimodal? - Answers Can be multimodal
