In this module we will see the problems with Goal Stack. We will also see how Plan Stake planning works and solves some problems cannot be handled by GSP.
Goal stack planning is simple method which works in recursive way but it is really hard to implement that in a manner which produces an optimal solution. Whenever a goal is achieved component by component, a solution of current state might disrupt previously solved components if the components have some inter-dependency. Sussman has shown that it is possible even for a very trivial blocks world problem that we do unnecessary stacking and un-stacking.
Plan space planning is a method of implementing planning in a nonlinear way. Actions are chosen to be added at any part of the plan. Actions are ordered based on one action being a cause of another action or some predicate being an effect of one action while it is in the prerequisite of another action. When an action can break the causal link or deletes or modifies the preconditions, the order is changed to make sure that such break does not take place. Though much more complicated, plan space planning can achieve far better optimized plan compared to goal stack.
|Problems with GSP and introduction to Plan Space Planning||00:31:00|
No Reviews found for this course.