Probabilistic "Events" in Robotics

You might be wondering what coins have to do with robotics.

A coin flip is a perfect example of a probabilistic event: a set of outcomes to some experiment where each outcome has a probability.

With a coin, the outcomes are clear: heads or tails, and the probabilities are simple: 0.5 and 0.5.

A self-driving car makes hundreds of calculations about probabilistic events every second, but the events are not as clean as a coin flip. For example:

• What is the probability that this sensor measurement is accurate to within 5 centimeters? What about 1 centimeter?

• What is the probability that some other vehicle will turn left at this intersection? Go straight? Turn right? What if they just sit there forever?

• The radar and lidar measurements seem to disagree! What's the probability that the range finder somehow became detached from the roof?

These examples are all much more interesting than "heads or tails?" but they are also less straightforward, which makes it much harder to learn probability theory from them.