In this session we will look at how to use particle filtering for localization of a robot.

Download the following file: particle.zip

It includes:

• Source code for a simple robot simulator
• A skeleton for a particle filter algorithm (ParticleFilter.java)
• The binary of a sample solution (sample_solution.jar)

• Implement the missing parts of ParticleFilter.java (use confusion=0.1 for the resampling)
• Experiment with the settings for noise, number of sensors and number of particles to see the influence on how fast the agent knows its position and how good the estimate is
• Try out what happens if you do not add white noise in the resampling step (set the confusion parameter of the resampling function to 0.0)

1. Your code
2. A few sentences about your observations and what you can conclude from them. How do the number of sensors, number of particles, noise parameters and the confusion parameter influence the accuracy of the position estimate? Why is that so?