Summary

In this chapter, we used the encoder data along with a model of our robot poses and an arena to create a dynamic probabilistic mathematical model of the robot's location. We learned how to mathematically represent the world the robot is occupying, and also represented possible poses the robot could be at.

We saw how encoder readings can infer where the robot has moved or turned, accounting for noise in the sensors, and then fused that with a model to observe whether the prediction is still within the arena boundaries. The robot has gone from simply reacting to sensors to making dynamic predictions and testing hypotheses about where it is.

We took a trip through probability, making random selections, weighting poses based on their properties versus the world, and saw how to display this on a browser while it runs. We saw how probability distribution concentration can be measured and used to control how often we resample from a population. These are powerful tools we can extend...