Better Neighbor A neighbor A'E N(o, A)resolves the inconsistency o by inserting or deleting an action Use evaluation function E(a) E(a=a Execution cost (a +β Temporal cost(a) +y Search cost(a) E(a)= E(a=aEXecution cost(a +β Temporal cost(a) +r Search cost(a) Relaxed plan Idea: Don't consider the mutex relation and perform reachability analysis Insert action a Find all actions that is required to support the preconditions of a Compute the maximum time duration required for all actions Return Set of actions added: Aset(EvalAdd(a)) Max time duration of the actions: End time(EvalAdd(a)) Remove action a Find all actions that is required to support all preconditions that were unsupported due to removal of a Compute the maximum time duration required for all newly inserted actions Return Set of actions added: Aset(EvalDel(a Max time duration of the actions: End time(EvalDel(a))9 Better Neighbor? • A’∈ N(σ,A) resolves the inconsistency σ by inserting or deleting an action. • E(a) = E(a)i = α⋅Execution_cost (a)i + β⋅Temporal_cost(a)i + γ⋅Search_cost(a)i E(a)r = α⋅Execution_cost (a)r + β⋅Temporal_cost(a)r + γ⋅Search_cost(a)r 10 Relaxed Plan • • – – – • • End_time(EvalAdd(a)) • – unsupported due to removal of a – actions – • • End_time(EvalDel(a)) A neighbor Use evaluation function E(a) Idea: Don’t consider the mutex relation and perform reachability analysis. Insert action a Find all actions that is required to support the preconditions of a Compute the maximum time duration required for all actions Return: Set of actions added: Aset(EvalAdd(a)) Max time duration of the actions: Remove action a Find all actions that is required to support all preconditions that were Compute the maximum time duration required for all newly inserted Return: Set of actions added: Aset(EvalDel(a)) Max time duration of the actions: