"Jeep problem and its 2D extension . the solution with an application of the reinforcement learning." by Marcin Plucinski
==================================================================================================
Reviewer: Akira IMADA, akira@bstu.by

1) How familiar are you with the subject matter of the paper? [0-5]: 4

2) Relevance to the conference [1-5]: 5

3) Contribution and originality [1-5]: 4

4) Paper readabilty, presentation and organization [1-5]: 3

5) Overall rating: 4

6) Comments to the Author:
First of all, the author should proof-read more carefully to avoid such an expression "(4) i (5)." There are still many such as those expressions.

It is interesting to challange the Jeep Problem by Reinforcement Learning.

Under his new interpretation, or condition, of the problem, i.e., (1) the amount of fuel for our disposal is unlimited; and (2) the goal is to cover the given distance d instead of searching for the maximum distance to be covered, the author concludes, "The presented method is a very effective tool in the exploration policy searching both for 1D and 2D case."

The reviewer, however, doubts it. In the standard version, our interest is how can jeep go (d) under a constraint that the jeep can refill (n) times. Assume now the capacity of the tank (z) = 15 to enable to apply Eq.1 for n = 3, then algorithm should find how jeep reach the maximum distance of d = 15 + 5 + 3 = 23 units.

So, author's experiment in which, for example, d = 6 for z = 3 under infinite n is still as tiny as [10] for which the author wrote "the task was solved there only in a limited scope."

The reviewer hope the author will develop this method to be more efficient, or "very effective tool" as author put it, to challange in a bigger scale.


7) Confidential Comments to the Editors:
None