Constraint-Procedural Logic Generated Environments for Deep Q-learning Agent training and benchmarking

While training and benchmarking neural networks, a large and precise set of data and an efficient test environment are parts of a successful process. A good data set is usually produced with high effort in terms of cost and human work to satisfy the constraints imposed by the expected results. In the first part of this paper we focus on the specification of the properties of the solutions needed to build a data set rather than using common primitives of imperative programming, exploring the possibility to procedurally generate data-sets using constraint programming in Prolog. In this phase geometric predicates describe a virtual environment according to inter-space requirements. The second part is focused to test the generated data set in a machine learning context by means of an AI gym and space search techniques. We developed a deep Q-learning model based neural network agent in Python able to address the NP search problem in the virtual space; the agent has the goal to explore the generated virtual environment to seek for a target, improving its performance through a reinforced learning process. © 2022 Copyright for this paper by its authors.