Repository for the most difficult AI assignment at the University of Umeå, Sweden.
The goal is to implement multiple algorithms to make a roomba-like robot explore an environment and map it using range-limited laser sensors. The robot computes frontiers to go to to discover the most unknown regions. I used an optimized version of A* on the configuration space to find paths to the frontier points. The laser sensor model uses Bayes theory to update the probabilities on a given cell of the map and other parameters such as previous readings, the range, etc. The path tracking algorithm (pure pursuit) used while travelling from a point to another originates from a previous assignment (in another of my repository: Umu-AI-Fundamentals).
The simulator used is MRDS, originally created by Microsoft but sadly not developed anymore. The assignment should transition to another environment in the future.
A video of an example run can be found here: https://www.youtube.com/watch?v=6mEpk-EFsFs