Grid and Monte Carlo

We will describe two localization algorithms that are capable of solving global localization problems.

• They can process raw sensor measurements. There is no need to extract features from sensor values.

• They are non-parametric.

• They can solve global localization and kidnapped robot problems.

The first approach is called grid localization. It uses a histogram filter to represent the posterior belief. The second approach is called the Monte Carlo localization, arguably the most popular localization algorithm to date. It uses particle filters to estimate posteriors over robot poses.