Option 1: Limit size of frontier, keep on only k best f(n) nodes Will be incomplete and sub-optimal Good selection of k helps make it fast, and potentially near-optimal Option 2: Keep all nodes with f(n’) > d f(n), where f(n) is current best
Backtracking Search