Development of a Shift Quality Metric for an Automatic Transmission

An automatic transmission shift quality metric has been developed which uses the vehicle's fore/aft acceleration and correlates it with various driver's subjective ratings.

Four components were extracted from the single fore/aft acceleration signal which relate to the human subjective response and vehicle resonance condition. These components were: peak-to-peak amplitude of a filtered acceleration, peak-to-peak jerk, maximum average power, and 10-14 Hz frequency content. They were then mapped using linear regression and a nonlinear neural network modeling technique to the various driver's subjective responses. Good correlation between the subjective ratings and the four signal components occurred for low shift disturbance when using linear regression mapping. Excellent correlation existed for all shift disturbances when using the nonlinear neural network mapping.

A computer program was written to automate the data collection and analysis. Test methods were developed for creating the subjective relationships and for benchmarking the shift quality of other vehicles.