Browsing by Author "Holmström, Axi"

Quantum Neural Networks (QNN) were used to predict both future steering wheel signals and upcoming lane departures for N=34 drivers undergoing 37 h of sleep deprivation. The drivers drove in a moving-base truck simulator for 55 min once every third hour, resulting in 31 200 km of highway driving, out of which 8 432 km were on straights. Predicting the steering wheel signal one time step ahead, 0.1 s, was achieved with a 15-40-20-1 time-delayed feed-forward QNN with a root-mean-square error of RMSEtot = 0.007 a.u. corresponding to a 0.4 % relative error. The best prediction of the number of lane departures during the subsequent 10 s was achieved using the maximum peak-to-peak amplitude of the steering wheel signal from the previous ten 1 s segments as inputs to a 10-15-5-1 time-delayed feed-forward QNN. A correct prediction was achieved in 55 % of cases and the overall sensitivity and specificity were 31 % and 80 %, respectively.