Just a decade ago, engine management was a relatively simple affair. Spark advance and fuel injection quantities were the key parameters, and the determination of both was a relatively simple matter of determining engine speed and load and using those values to index a lookup table. Nowadays, however, technology to improve power, drivability, and fuel consumption - such as variable valve timing and lift, intake tuning, and variable displacement - have greatly complicated the task of calibrating an engine control unit (ECU).
Steve Magner from Ford was at the 2006 SAE Convergence electronics show to speak on this topic and discuss what steps might be taken in the future to improve engine performance and decrease time-to-market. He used the term "high degree of freedom" (HDOF) to describe the typical modern powerplant, and explained that the table-based method ("full factorial mapping") of calibrating ECUs has basically reached its limit. Where as it's desirable to complete the calibration of a new engine in about four months, a HDOF engine currently requires up to 15 months to fully define in the laboratory. In one example, developing just the timing maps for the EPA' s fuel economy test cycle involved 630 states of operation and three months of time on the dyno. Design-of-experiment techniques can be developed to simplify the mapping process, but at the cost of effectiveness - over 25% of a new technology's potential benefit can be lost via incomplete mapping efforts.
Ultimately, the solution proposed by Magner is the direct monitoring of combustion quality, with closed-loop feedback constantly working to optimize the various operation parameters. The two technologies currently being explored are pressure measurement and ion detection schemes, with the former being easier to implement but the latter perhaps being more cost-effective (indeed, Saab has used such a technique for quite some time, albeit with significantly reduced functionality than what is being discussed here). Variations between each cylinder will require that each carry its own sensor, and the required resolution (with data collected for each degree of crankshaft rotation) requiring computational horespower beyond that provided by modern ECUs. Systems based on digital signal processors (DSPs) may be of some use to make sense of all this data and react quickly enough during transients.