Påhittade tentafrågor

Question 1

Maximum Likelihood (ML)

Answer 1

Estimates the parameter that makes the observed data most likely. Maximizes P(data | parameter). Does not use prior knowledge.

Question 2

Maximum A Posteriori (MAP)

Answer 2

Estimates the parameter using Bayes’ theorem and a prior. Maximizes P(parameter | data). Useful when prior information is available.

Question 3

Taylor approximation in GPS

Answer 3

Used to linearize pseudo-range equations that include square roots. Helps solve the system using least squares and iteration.

Question 4

Dead reckoning

Answer 4

Estimates position using the robot’s own motion data (e.g. encoders, IMU). Prone to drift over time without external references.

Question 5

Visual odometry

Answer 5

Estimates motion by analyzing changes in camera images. More accurate than dead reckoning but affected by lighting and visibility.

Question 6

Pulsed radar with missing echoes

Answer 6

Still gives range to static objects, but motion tracking becomes harder. Errors can be due to angle, material, low reflectivity or multipath.

Question 7

Multipath effect

Answer 7

Occurs when a radar signal bounces off multiple surfaces, lika buildings, before returning, causing delay or false readings.

Question 8

Motion capture system

Answer 8

Uses external cameras and markers to track pose. Very accurate, works indoors, needs clear line of sight, and struggles in bad lighting.

Question 9

Visual odometry – error sources

Answer 9

Fog, dirt, low texture, fast motion, and lighting changes. Can be reduced by using IMU, good lighting, or preprocessing.

Question 10

Kalman filter

Answer 10

Combines prediction and correction to estimate position over time. Used in sensor fusion to weigh multiple sensor inputs.

Question 11

Sensor fusion

Answer 11

Combining data from multiple sensors to get a more accurate and reliable estimate.