AI Chef - Analysing a Planning Problem

This report details two different approaches for solving the ‘Child-Snack’ problem. The first approach was to tackle it as a PDDL problem formulation and the second approach was to reformulate it as an optimization problem. To solve the PDDL problem formulation Breadth-first, Serialized Iterative Width+ and DFS+ were used. To solve the optimization problem, AMPL was used together with IBM’s Ilog cp solver. This optimization version of the problem turned out to be tedious to formulate and difficult to solve.

The report contains brief descriptions of the algorithms used for the solutions and an analysis of the results. The results show that a PDDL approach is more efficient, both in terms of implementation and the algorithms’ search for optimal solution, but the solvers used with that approach did not give as good paths, in terms of path length, as when it was formulated as an optimization problem in this project. The code can be found in this github repo.

Full Report