• Strategy: expand a cheapest node first
  • Implementation: fringe is a priority queue (priority: cumulative cost)
  • Properties: (if the solution cost C* and arc costs at least 𝜺 β†’ effective depth is roughly C*/𝜺
    • Time complexity: \\(O(b^{C*/𝜺})\\)
    • Space complexity: \\(O(b^{C*/𝜺})\\)
    • Complete: yes (if C* and 𝜺 are finite and positive)
    • Optimal: yes
  • pic
  • Good?
    • complete & optimal
  • Bad?
    • explore options in every direction
    • no info about goal location