Probably a stupid question but a thought did come into my head about analysis of the MoTeC data. Has anyone tried uploading a MoTeC file to an AI (not the computer cars) provider such as ChatGPT etc to see where improvements can be made? I know nothing about these but assuming there is sufficient data for it to learn from I am assuming that AI could provide an analysis? How good it would be I have no idea but as I said just a thought.
Sorry, a little OT. This is what I had on my old steering input overlay on Simhub: Code: [DataCorePlugin.GameRawData.CurrentPlayerTelemetry.mUnfilteredSteering]*[DataCorePlugin.GameRawData.CurrentPlayerTelemetry.mPhysicalSteeringWheelRange]/2 This is the modified version for LMU and it's just ok comparing to my real wheel position Code: [DataCorePlugin.GameRawData.CurrentPlayerTelemetry.mUnfilteredSteering]*[DataCorePlugin.GameRawData.CurrentPlayerTelemetry.mVisualSteeringWheelRange]/4
ChatGPT wouldn’t be any good at this type of analysis. The way to think about ChatGPT is that it is guessing the next word in a sentence. It doesn’t understand numbers. The way for algorithms to understand how to get faster around the track would be to develop some sort of reinforcement learning algorithm. So as the algorithm completes more laps it better learns the track. There was a motorcycle game that took this approach. It would be interesting for sim racing because you could theoretically have it optimize setups as well. You could also have it learn to race with other cars, but would have to have a really smart penalty system or else it would probably learn to put competitors in the wall with a little bump to the tail. The compute required would be enormous.
