Experimental hyperc-defined PDDL planner
Because the planning problem is defined as the planning problem itself - the system is suitable for searching for a universal, domain-independent heuristic with machine learning.
Several planner features are targeted:
- support infinite level of recursion of itself to allow searching for a better heuristic of searching a better heuristic...
- definition-modifying actions to optimize the problem itself as part of solving process
- same actions to be used as auto-fixing for incorrect problems which shoud make all problems solvable
The project is under heavy development. If you are interested in more details please write at andrew@hyperc.com.