The research on the improved light version max restricted path consistency algorithm with the learned value ordering heuristics.