Auto-Tune Hardware with Accelerometer Feedback

Is it possible to program a utility that will allow X-Sim to self-tune with the hardware?

With so many different configurations and design approaches, it is up to each builder to determine all of the necessary settings to be sure that the hardware produces the correct motion cues for the driver/pilot. Once the basic setup is complete and the syntax is correct, X-Sim is ready to drive the hardware. At this point, the system is capable of moving the cockpit/seat into all possible positions, but it still does not know what action results in the pilot sensing acceleration, braking, lateral forces, etc. To achieve this, we modify dozens or hundreds of parameters until it feels right. Instead of trial and error, why not use sensors to tell X-Sim exactly what forces the pilot really feels?

Let's say we mount a pair of accelerometers into the driver's seat, one low in the seat, and one up high near the driver's chest or head. This arrangement would be standardized such that proper location of the sensors results in the measurements of one simulator equaling those of another (given the same seat position on both rigs). We then run a calibration routine in X-Sim; this drives all of the hardware through the full range of motion, during which it records the acceleration forces felt by the sensors. By comparing the measured cues to those of a known calibration, the software now knows what the hardware is capable of simulating. At this point, all of the configuration parameters that everyone would normally have to develop or borrow from others is done automatically. The system is now self-tuned.

Thoughts?

The concept is great, but how are you going to implement it?

Do you have a proposed solution or is this something that just came across your head?

In practice it is much more complicated since a lot of acceleration forces are simulated by tilt and position.
A platform cannot accelerate only for a short period of time before it reaches it's limits...

It would be LOTS and LOTS of work, and in the end you might still end up with poor motion quality.

...and for certain type of platforms you need Gyroscopes, NOT accelerometers (i.e. Joyrider!!)

And something else, accelerometers are usually measure acceleration in static, that means you will have measurement of the sudden start of motion and the sudden stop... I know this from first hand since I was trying to make a segway with only accererometers... hahaha!!

Correct, Thanos. We actually need to use a gyro to do this.

The Plan:
1. Mount a gyro onto seat or frame, properly aligned with x,y,z axes of motion.
2. Connect gyro to pc to retrieve real-time data for roll, pitch, and yaw.
3. Run calibration utility (yet to be developed). Calibration utility commands simulator through full range of motion while recording speed of movement, range of motion, settle time, etc. A hardware capability profile is generated for the system.
4. Save hardware capability profile to X-Sim.
5. X-Sim uses hardware capability profile data for simulator to execute the appropriate commands for best simulation with current hardware.

We need to find a gyro that is accurate enough for the task, but inexpensive enough for everyone to add to their system. Note that this concept is for setup, calibration, and optimization; I am not suggesting this as a means of control feedback for normal operation. This is a temporary closed loop feedback system used to measure and record the parameters for open loop control of the simulator.

Benefits:
1. New builders progress from basic hardware control to good or excellent motion simulation in record time. Several steps are either eliminated or greatly simplified.
2. Community testing and feedback results in an optimal hardware profile that produces the most realistic motion cues possible for a given system. If one builder's rig can pitch up 30 degrees in 0.05 seconds, then their calibration result will indicate this capability and X-Sim can scale this as necessary for the desired result. If another design is 10 times slower, then X-Sim uses this as a baseline for that user.
3. User feedback results in more advanced motion profiles resulting in better motion simulation results. These are deployed via the calibration routine; users run the new routine and generate the more advanced profile already tuned for their system.

This is an interesting idea, and one that I have been pondering / toying with as well.

One of the difficulties of implementing this, is that X-Sim provides for alot of flexibility in sim design. In addition to your hardware capability profile, you would also need to make calculations for each sim style.

A seat moving sim would need completely different calculations from a joyrider, which would need different calculations from a 6DOF, which would need different calculations from Nimas sim design, which would need different calculations from my upcoming sim design

I don't think there can be a one size fits all calculation which would mean:
-Hardware capability profile
-Calculation plugins by simulator type
-Most likely still need a motion profile to optimize movements for game / vehicle

It seems like alot of added work for a pretty small gain in initial setup. Known simulator designs already have a variety of motion profiles available that are scaled correctly and I'm pretty sure this mechanism won't help much with unknown or new designs.

Maybe if you can come up with the formula for a universal motion cue the idea could fly. Am I missing something?

EDIT: Perhaps this could be useful in the case of the JoyRider design because everyone uses a different motor, controller and sim dimensions. A motion profile on one JoyRider probably does not translate well to another. In the case of SCN5 sims, the need is not really there because the motors / controllers are fixed, as are the reasonable dimensions, therefore a motion profiule translates well from one sim to another.

It all sounds very good!

What you are describing requires countless hours of programming, interfacing and hardware design. There are a lot of complications, but nothing impossible.

First, I think the community has to grow. There is just not enough users and not enough developers.

There is also need for hardware that is compatible with such a tuning mechanism. I just cannot see it being automated without extra intelligence of the motion controls.

In the end, it might just be easier to learn how to tune your platform manually.... Why would we take the best part of a DIY simulator away?

It appears that I have tried to oversimplify the motion cues that are created by the different designs that are in use. I just realized (via other posts on this forum) how different the motion cues are for the frex sim compared to the joyrider sim. Obviously the sensation of the movements will be slightly different, but the specific interaction of the sim with the driver is so fundamentally different between these two designs. I saw others discussing cascaded systems where a seat mover is placed onto a 6DOF platform, or some variation of this. In the real world, the car moves around in response to the driver's input and the vehicle's interaction with the track; inside the cockpit, the driver moves around in the seat based on the current position and movement of the vehicle. It would seem logical that the best way to represent this in a sim that is limited to a finite space would be to nest two or more systems together. Now I just have to decide if I really want to tackle something this huge on my first build. Probably not the best idea. But, when we do get to this level as a community; then it will be time to develop the auto-tune program. With a 2DOF + 6DOF working together, it's going to take a lot of calculating, testing, and adjusting to get these two systems to complement each other properly.

For your first build I recommend starting with something simple. Believe me, it will not be that simple!

When you finish it, you will have a better idea of what can be done and what is worth doing.