Preliminary Results on Reducing the Workload of Assistive Vehicle Users: A Collaborative Driving Approach

Abstract

Nowadays, physically impaired people still struggle with daily tasks when using mobility aid devices, whether for crossing doors, parking or manoeuvring in their homes. In this context, assistive robotics can offer solutions to those problems, thus increasing the users’ quality of life. However, studies must be performed to determine the best architecture for human–robot interaction. In this work, we propose a collaborative navigation strategy for improving users’ skills for driving assistive vehicles. We present four navigation modes: manual, assisted manual, autonomous and assisted autonomous. In particular in the two assisted modes, the system is able to predict the user’s motion intentions, reducing his/her workload. The system was validated in a real world environment with a population of twenty volunteers. Objective and subjective metrics were used to asses the system’s performance and usability, with special consideration to human factors. Results show that the system aids users to perform navigation tasks in a clear and compliant manner using a robotic assistive vehicle, while decreasing their perceived workload by 15% for the assisted manual, 41% for the autonomous and 40% for the assisted autonomous, when compared to the manual mode. Additionally, it is shown that if autonomous navigation sets a lower bound for user workload, the system approximates this bound while improving performance.

Appendix: Modified SWAT Questionnaire

To assess users mental workload, a modified SWAT questionnaire and methodology was used. The original SWAT questionnaire separates mental workload into time load, mental load, and stress load experienced by the person, and associates three degrees of intensity for each of these loads. Before commencing experimentation, SWAT requires users to order the 27 possible combinations of time load/mental load/stress load from the ones that produce them the least mental workload, to those that produce the most. The order of each combination is used to later scale every users’ perceived loads and associate them with a quantitative measure of mental workload. However, this sorting activity can take quite a long time, and thus for this work it was modified into a questionnaire to be answered previous to beginning the experiments. The format is as follows:

In what degree do the following factors affect your perceived mental workload? Evaluate from 0 (no influence) to 10 (much influence).

• Time load Time required vs time available to perform a task, temporal pressure to perform a task, having to distribute time between several tasks, etc. $\alpha _{TL}$ : [Value from 0–10]

• Mental load Complexity of the task, perceived difficulty, multiple decision-making, performing calculations, memory use, performing estimations, etc. $\alpha _{ML}$ : [Value from 0–10]

• Stress load Fatigue, motivation, emotional stress, uncertainty regarding security and safety, probability of failure, tension, perceived performance, etc. $\alpha _{SL}$ : [Value from 0–10]

This maintains the same idea of creating a relative scaling function between the three type of loads but greatly accelerates the process. After each experiment is carried out, the original SWAT requires users to evaluate their perceived time load, mental load, and stress load, with three levels available for each one. Later, using the scale based on the original sorting process, the selected values of each dimension of mental workload generate a final number ranging from 0 to 100 (or 0–1), where 0 indicates minimum workload, and 100 maximum perceived mental workload. For the modified version used in this work, the range is extended from 1–3 to 0–10 for each dimension, and the chosen values are scaled using $\alpha _{TL}, \alpha _{ML} ,$ and $\alpha _{SL}$. This addresses one of the most significant weaknesses of SWAT: low resolution when comparing activities with similar mental workload (as is the case with the navigation modes in this work). The format is as follows:

Please assign a value from 0 to 10 corresponding to the perceived presence of each factor of mental workload, during each navigation task.

• Time load 0–3: There is often spare time. Interruptions or overlap among tasks or activities occur infrequently, or not at all. 4–7: Occasionally there is spare time. Interruptions or overlap among tasks or activities occur frequently. 8–10: There is hardly any spare time. Interruptions or overlap among activities are very frequent, or occur all the time. SL : [Value from 0–10]

• Mental load 0–3: Very little concious mental effort or concentration is required. The activity is almost automatic, requiring little or no attention. 4–7: A moderate level of concious mental effort or concentration is required. The complexity of the activity is moderately high due to uncertainty, unpredictability, or unfamiliarity. Considerable attention is required. 8–10: Extensive mental effort and concentration is necessary. Very complex activity requiring total attention. ML : [Value from 0–10]

• Stress load 0–3: Little confusion, risk, frustration, or anxiety exists and can be easily accommodated. 4–7: Moderate stress due to confusion, frustration, or anxiety noticeably adds to workload. Significant compensation is required to maintain adequate performance. 8–10: High to very intense stress due to confusion, frustration, or anxiety. High to extreme determination and self-control required. SL : [Value from 0–10]

Finally, these answers are scaled to generate a value from 0 to 1, which serves as a metric for the perceived mental workload for each navigation mode. This workload index is calculated as follows:

$$\begin{aligned} WL = \frac{TL \times \alpha _{TL} + ML \times \alpha _{ML} + SL \times \alpha _{SL}}{10 \times (\alpha _{TL}+\alpha _{ML}+\alpha _{SL})} \end{aligned}$$

(4)