Cognitive workload-matched haptic feedback for automated driving takeover: A two-stage physiology-informed Wizard-of-Oz simulation study

Fixed-parameter haptic warnings may perform differently across cognitive workload states, while adaptive HMI studies often confound workload-matched strategy effects with specific haptic-parameter effects. We conducted a two-stage physiology-informed Wizard-of-Oz (Physio-WoZ) driving-simulator study of workload-matched haptic feedback for automated-driving takeover. Experiment 1 used a 2 × 4 within-subject design to screen No Warning, High-Gain Continuous, Low-Gain Pulsed, and Neutral Pulse under low and high workload in 40 valid participants performing manual car-following emergency braking. Experiment 2 used a 2 × 4 within-subject takeover design with 60 valid participants to evaluate No Warning, Static High-Gain, Static Low-Gain, and Load-Matched Adaptive strategies. Among the tested configurations, High-Gain Continuous facilitated attention reorientation and increased perceived urgency under low workload, whereas Low-Gain Pulsed produced faster braking responses, higher minimum TTC, lower steering angular velocity, and lower annoyance under high workload. Strategy validation showed that the Load-Matched Adaptive advantage was concentrated under high workload: relative to No Warning and Static High-Gain, it produced shorter takeover time, faster braking response, higher minimum TTC, lower SDLP, and higher trust; relative to Static High-Gain, it also reduced the near-collision rate. In both experiments, EEG, GSR, and HRV supported workload-manipulation checks and offline validation; haptic output followed predefined workload labels and experimental conditions rather than an online physiological classifier. These findings support workload-matched haptic feedback as a complement to fixed-parameter warnings, while Physio-WoZ provides a controlled validation pathway before real-time closed-loop deployment.