5

The concept of "planning" is not just related to RL. In general (as the name suggests), planning consists in creating a "plan" which you will use to reach a "goal". The goal depends on the context or problem. For example, in robotics, you can use a "planning algorithm" (e.g. Dijkstra's algorithm) in order to find the path between two points on a map (given e....


3

Welcome to AI.SE @Israr Ali. The problem of scheduling a timetable is an example of a constraint satisfaction problem, a topic long studied in AI. There are many possible techniques to apply to this kind of problem. They can be organized into three broad categories: Global search algorithms, like backtracking search can be used to try and find an ...


2

The relationship between search and planning is not made clear in the material you are reading, which is one of many reasons MIT doesn't use AI by Russell and Norvig anymore. It isn't the modern approach now and wasn't in the best survey course textbook when it came out, but it was politically correct with the faculty then. An online and updated version of ...


2

Not all search is planning (is A connected to B), but all planning is search (how do I get from this to that). Here's an example in Prolog with a domain described in terms of actions, when they are possible, and what the result of the actions are. The description is of an uncomputed graph of un-calculated size where each node is a situation and each edge ...


2

I am aware that as the experience set grows the Central Limit Theorem will come into play and the distribution of experience will more accurately represent the true environment's state-actions-rewards distribution I believe here you mean the Law of Large Numbers which states that for a large enough sample ($n \rightarrow \infty$) the sample mean will ...


2

By far the most commonly used strategy is to select the child with the highest number of visits. This is as described in the 2008 paper you linked. It's also what's referred to as the "robust child" in the 2012 paper you linked. In algorithm 2 of the 2012 paper, they actually use the highest average reward, which corresponds to "Max child". It looks like ...


2

Is this the only big difference? Or in other words, can I do the same implementation as in a simple MCTS with the 4 stages: selection, expansion, simulation and backpropagation, where the result of the simulation is the accumulated reward instead of a value between 0 and 1? How would the UCT selection be adjusted in this case? No, this is not the only ...


1

I don't know of any work on this with respect to PDDL, but this is very similar to a conceptual dependency application called SAM (Script Applier Mechanism). Conceptual Dependency (CD) models actions using a number of primitives (which could be seen as equivalent to PDDL primitive actions): PTRANS for physical transfer, PROPEL for application of a physical ...


1

Ah hah! The way I had defined the disks made d5 the LARGEST disk, not the smallest. So, the last few lines of the file should be: (clear d4) (clear d2) (clear d5) (on d4 peg1) (on d1 peg2) (on d2 d1) (on d3 peg3) (on d5 d3)) (:goal (and (on d1 d2) (on d2 d3) (on d3 d4) (on d4 d5))) ) However If I wanted the opposite to be true, d1 to be the ...


1

First place to look is how the preconditions/effects of different actions interact.


1

You have stumbled upon a common drawback of the vast majority of modern planning technology. The "flattening" you refer to is actually called "grounding" in the community. Indeed, grounding is the first step of almost every planner out there. Planners that don't do this grounding phase are typically referred to as "lifted planners", but their availability is ...


1

The question doesn't really make sense: PDDL is a description language that is used to formulate a problem. This description then is the input to a planner; how the planner arrives at the intended solution is not related to the PDDL description. There are a number of planning algorithms, and you can implement any of them to make use of a PDDL description. ...


Only top voted, non community-wiki answers of a minimum length are eligible