-
Notifications
You must be signed in to change notification settings - Fork 17
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[IDEA] - Crowd Source AI #5
Comments
Hi @mintuz I am currently working on an idea to use a reinforcement learning algorithm ( https://github.com/karpathy/reinforcejs ) where we use the velocity, distance traveled and heading from BB8 as inputs. The output will be the actions roll forward, backwards, turn left or turn right and based on collision detection, feedback a reward to the RLA so that it can adjust to behaviors where it avoids collisions and hopefully maps out a room. The issue I am facing is where OSX is unable to get a response back from BB8 once a command is issued:
My current environment is:
Do you have any inputs towards this? |
Hey this is a great idea, really look forward to see what you can achieve with it as for the issue you are seeing, I've not seen this before but can take a look for you at the weekend if you still haven't got anywhere with it by then. My initial thoughts though may be a compatibility issue with the Sphero API and the version of node you are using as that's the only difference between your machine and mine. I am running node It may also be to do with how you are running BB8 Commander, I haven't used it as a library in an external project for a while and it was a really basic project just to test the functionality. Maybe there's a bug in there somewhere. I will certainly take a look at the weekend though if you haven't managed to figure it out. If you do find out if it's an issue with BB8-Commander itself please do raise a PR. |
Thanks for looking into this, I actually have not tried this with the Commander module yet, I am having issues with the vanilla SpheroJs module. I will give it a try with the Commander module and let you know how it goes. As for the reinforcement learning portion, I believe the temporal limitations of the DQN algorithm might limit BB8 from learning to map a room as the steps taken to get a negative reward will be too far in time between positive rewards. In any event we will have to cross that bridge when we get there! UPDATE: I tried the BB8 Commander CLI module and received the following error:
|
Let someone else control your BB8 when you can't. We then AI!
The text was updated successfully, but these errors were encountered: