Talleres teóricos

Each test subject must complete one test run consisting of a track with a length of 32 km. The constant speed is 60 km/h, which results in a total test length of 32 minutes. The test subjects drive on a rural road with a lane width of 3.75 m and mainly easy curves with a radius from 150 to 700 m. They are always on the priority road with no need to stop at intersections. During the drive, the participants always drive with the automation switched on and at least one hand loosely on the steering wheel. There is no traffic on the road in the direction of the test vehicle, but frequent oncoming traffic in the left lane. We did not introduce a leading vehicle ahead of the ego vehicle in order to avoid potential negative effects of the carpet overlaying the vehicle ahead, even though it would have been reasonable to make the limited speed plausible despite a free road. With the automation turned on, the virtual vehicle is kept in the centre of the right lane

Figure 5.15: Left: Control buttons on the steering wheel, the I/O-key (bottom left) toggles longitudinal and lateral automation. Right: Digital instrument cluster, displaying current speed and status of the automated longitudinal and lateral control system.

and follows a pre-computed path. In case of an error this path deviates away from the lane centre to the left, towards the left lane and continues to deviate with an increasing offset (as described in 5.1.4). This happens on seven locations on the track. In this case, the test subject must recognise this incorrect behaviour and overrule the steering wheel by turning it in the opposite direction. The test subject must then manually bring the vehicle back on the centre of the right lane and switch the automation on again. The errors have all been placed in curves with a radius of 300 m for comparability. However, the curves have been embedded in a changing surrounding environment so that it cannot be foreseen if the next curve will produce an automation error. We designed the error situations in a way that not overruling the steering wheel would result in a collision with the oncoming traffic. The oncoming vehicles have been programmed to control their speed so that they would hit the test vehicle when entering their lane. With that we prevent any ambiguity and avoid that the test subjects accept the deviation from their lane and react only when they would leave the road. We decided to place the error situations only in curves after having a test subject close his eyes before an error occurred on a straight section of the road. He could tell immediately when the error became perceivable by the slightest movements of the steering wheel. When placed in a curve, the steering wheel naturally turns and the error cannot be anticipated. Three errors occur in a right curve and four in a left curve. The first error occurring is treated separately as theinitial contact error. The errors are located as follows:

1. Left curve: Initial contact error (after 5.48 minutes) 2. Left curve (after 7.50 minutes)

3. Left curve (after 13.50 minutes) 4. Right curve (after 16.01 minutes) 5. Left curve (after 23.23 minutes) 6. Right curve (after 24.55 minutes)

7. Right curve (after 31.01 minutes)

We applied a factorial between-subject-design, dividing our test sample into three groups according to the carpet version: Contact analogue head-up display (HUD, cf. Fig. 5.13, left), extra in-car display (DISPLAY, cf. Fig. 5.13, right) and aBaselinedrive with no visual feedback at all.

Before the test drive there was a short demonstration drive during which we made the test subjects familiar with the driving simulator, the control, the overruling of the automated system, and if applicable, with the characteristics of the driving path display. They were instructed to keep the automation on all the time, but if they thought, something was going wrong, they were supposed to take over manual control. We told them their primary task was to make sure that the vehicle drives safely on the road. They werenot shown and told how an automation error finds expression, in order to obtain genuine data on initial contact with an error.

Additionally there were two secondary tasks to accomplish that were selectively presented during the test drive: entering destinations into the navigation system using the iDrive controller

(navigation) and answering questions in a phone conversation (phone call). These tasks were

implemented identically as described in section 3.1.4. The tasks were chosen based on the results from the ALCT studies (cf. 4.2.5 and 4.3.4). We decided to include two active tasks, one visually and the other auditorily demanding that had also produced a significant effect in the ALCT setting. Each secondary task was performed once during an error in a left and in a right curve and randomly distributed over the locations with exception of the initial error location, which always occurred without a secondary task. In order to prevent predictability, each task had to be performed two more times placed on predefined locations on the road without an error occurring. Appendix Table A.3 shows the permutation table for all tasks in 12 different variants. Below follows an overview of the relevant factors in this study:

• _{Carpet versionwith the levels:}

– Contact analogue head-up display (HUD)

– Extra in-car display in the centre console (Display) – No visual support (Baseline)

• _{Type of curvewith the levels:}

– Left curve – Right curve

• _{Type of secondary taskwith the levels:}

– Phone call – Navigation – No task