No Subject

[Mike Klopfer]

Looking at the mail on this mailing list I get the idea that some of the
posters are working together to develop a fi controller.
I'm wondering if there is an ftp site or www server that has information about
the overall architecture of this fi system being designed, perhaps a list of
tasks needing to be done etc. I'm running linux so I could compile some cross
compilers or possibly the kernel that was mentioned. I didn't think gnu had a
compiler to produce 6800 assembly though.
Also I could probably design A/D circuitry to interface to the CPU. The biggest
unknown in my mind is how I'm going to come up with a control algorithm once I 
get all the rest of the system. Perhaps there is a good general purpose 
solution to this problem. If so I'd appreciate information about where to look. 
I don't know if I just need to come up with values to parameters of some 
general purpose control equation. I imagine this is wishful thinking at least 
if I want an efficiently running engine.

I've been thinking along the following lines for designing an "optimal"
engine control algorithm. Suppose you take an existing engine controller
and record the sensor and actuator (injector pulse width, dwell, advance
etc) variable values for a period of time. Using this data you develop a
model of the engine state space (a function which predicts the sensor values
at the next sample time as a function of previous sensor and actuator
values). Are there already accurate (and hopefully simple) models of engine
response to actuator input? I read a fuel injection book by Bosch that had a 
picture of experimental data showing the output of some engine parameter as a 
function of some input parameters which makes me think that it is likely that
there aren't any such models covering all engines. If there is such a model
the data can be used to determine any fixed parameters of this model. Then
you can develop an optimal control equation based on this model. You can
then run the engine with this control equation occasionally perturbing the
control to see if the engine responds according to the model. If not the
model can be refined which can be used to obtain new control equations.

I suspect that the above description is unclear but I guess I'll see what
kind of response it generates before elaborating. Thanks.

mike