Lecture 7 - Occupancy Grids Flashcards
What’s the Pose of the robot?
The rotation and translation of a robot, or in general its full state configuration.
What’s the Odometry of the robot?
The transformation of the body frame with respect to its initial pose (fixed frame of reference).
What’s the Dynamics Model of the robot?
What is the next state given current state and control?
What’s the Sensor Model of the robot?
What value will be returned on a given sensor channel from a given state?
Why is mapping a problem?
Because building maps is hard even for humans.
What are the two main sources of uncertainty in robotic mapping?
- Accumulating uncertainty in the dynamics.
- Uncertainty in sensor measurements.
EXTRA: maps change.
State the probability of a robot’s state.
p(xt+1 | xt, ut)
State the probability of a robot’s sensor reading.
p(zt | xt)
State the probability of any map
p(m | x1:t, z1:t)
Why is the probability of a map that considers all previous paths and sensor readings instractable?
for a 100*100 grid there are 2^10000 possible binary maps.
What probability approximation makes mapping easier?
Approximation ignores all dependencies between map cells, given known info. Assumes (for tractability) that cells are independent given path and measurements.
Explain the steps in Inverse sensor measurement model
Given map cell (i,j) , the robot’s state xytheta , and beams rk
Find index k of sensor beam that is closest in heading to the cell
If the cell is sufficiently closer than rk
// Cell is most likely free
Return that is well below 0.5
If the cell is sufficiently farther than rk or out of the field of view
// We don’t have enough information to decide whether cell is occupied
Return prior occupation probability
If the cell is nearly as far as the measurement
// Cell is most likely occupied
Return that is well above 0.5
What’s the threshhold for the maximum likelihood map?
p = 0.5
