0 votes
asked by (12 points)

Hi,

I am using X-Plane for an academic research project to attempt to teach an unsupervised agent to learn to fly an aircraft. To do this I was planning on using reinforcement learning to allow the agent to try different things to figure out how to fly the aircraft by itself.

For anyone that has done something similar will be aware of the amount of time it takes an agent to "learn" to do something as well as the amount of computer power required. To this end I had 2 questions:

1) Is there a version of X-Plane that has only the core simulation frame work that it uses for flight simulation. What I mean by this is there a version that can be run without the interface and that can just be controlled through the API. The reason for this is I would like a version of the simulation module that can be run and controlled on a large High Performance Cluster which are  command line driven and have no graphical interface support.

2) Is it possible to run multiple versions of X-Plane side-by-side to allow for multiple agents to run small modifications across multiple simulations at a time to allow the overall agent to apply NEAT principles across multiple networks to improve the ability of the agent to fly the plane in a shorter time.

Any advice or information in this matter would be greatly appreciated.

Many Thanks,

Matt

Please log in or register to answer this question.

Welcome to X-Plane Q&A, where you can ask support questions and get answers from members of the community.

This site is for X-Plane support questions only. Please search for existing answers before posting your question. Off-topic questions will be locked.

If you’re new, you’ll need to register before asking your first question.

If your question is answered, click on the check mark to select the best response.

Categories

...