Heuristic Details Sample Clauses

Heuristic Details. ‌ As the planning task spans a graph by applying a selection of set of rules, the planning task abstraction is generated by abstracting the initial state, the goal state and the operators (modifying them according to the pattern in use). Important to note is that, while plans in the original space have counterparts in the abstract space, the opposite doesn’t always hold true. This is important, as the main idea of the abstraction heuristics are that it is easier to find a plan by applying uniform blind search from φ (s0) to φ (s∗) in the abstract state-space A , compared to searching from s0 to s∗ in the concrete state-space P. Looking at Figure 2.9, we can see the same abstraction on two problems defined in the office domain. The first one, Figure 2.9a represents the abstraction on the problem defined initially, where Xxxxxxx, the robot assistant can move between all locations and has the transition diagram present in Figure 2.4. The second one, Figure 2.9b represents a slightly modified problem, where there is no path for Xxxxxxx between wp2 and wp3. Both planning tasks will have the same abstract state-space. PDBs are an admissible, consistent and additive heuristic. We can prove the first property as the lookup table consists of the optimal solutions in a relaxed state space. The consistency property is trivially shown by knowing that the shortest path distances satisfy the triangular inequality [29, 54].