Adaptive driver-automation shared steering control via forearm surface electromyography measurement
Zheng Wang, Satoshi Suga, Edric John Cruz Nacpil, Zhanhong Yan, Kimihiko Nakano
Abstract — Shared steering control has been developed to reduce driver workload while keeping the driver in the control loop. A driver could integrate visual sensory information from the road ahead and haptic sensory information from the steering wheel to achieve better driving performance. Previous studies suggest that, compared with adaptive automation authority, fixed automation authority is not always appropriate with respect to human factors. This paper focuses on designing an adaptive shared steering control system via sEMG (surface electromyography) measurement from the forearm of the driver, and evaluates the effect of the system on driver behavior during a double lane change task. The shared steering control was achieved through a haptic guidance system which provided active assistance torque on the steering wheel. Ten subjects participated in a high-fidelity driving simulator experiment. Two types of adaptive algorithms were investigated: haptic guidance decreases when driver grip strength increases (HG-Decrease), and haptic guidance increases when driver grip strength increases (HG-Increase). These two algorithms were compared to manual driving and two levels of fixed authority haptic guidance, for a total of five experimental conditions. Evaluation of the driving systems was based on two sets of dependent variables: objective measures of driver behavior and subjective measures of driver workload. The results indicate that the adaptive authority of HG-Decrease yielded lower driver workload and reduced the lane departure risk compared to manual driving and fixed authority haptic guidance.
Index Terms —Haptic guidance, driver-automation shared control, adaptive automation design, driver workload, surface electromyography I. Introduction
RIVING automation has drawn much attention in recent years, and remarkable outcomes have been attained with adaptive cruise control and lane keeping systems. However, before fully automated driving cars can be introduced into the market, keeping the driver in the driving control loop with a sense of agency is still essential [1, 2]. Shared steering control, which combines the abilities of human driver and the vehicle automation, has been developed as a suitable approach to keep the driver in the steering control loop, while reducing driver workload [3, 4]. Ideally, when driving with a shared steering control system, the driver can comfortably rely on haptic guidance torque to drive more safely. It has been found that haptic guidance can assist drivers with curve negotiation by producing proper
This work was supported by a Grant-in-Aid for Early-Career Scientists (No. 19K20318) from the Japan Society for the Promotion of Science. Z. Wang is with the Institute of Industrial Science, The University of Tokyo, Tokyo 153-8505, Japan (corresponding author: 81-03-5452-6916, e-mail: [email protected]). S. Suga, is with the Institute of Industrial Science, The University of Tokyo, Tokyo 153-8505, Japan; Technical University of Darmstadt, direction and torque on the steering wheel [5]. When drivers have low attention caused by fatigued driving, the haptic guidance steering reduces their lane departure risk [6]. When visual information from the road ahead is degraded, as in the case of dense fog, haptic guidance steering yields better lane-following performance than manual driving [7, 8]. A haptic guidance system has also been designed to assist drivers with prompt and steady steering for emergency obstacle avoidance [9]. On the other hand, it is suggested that fixed automation authority is not always appropriate with regard to human factors, in comparison to adaptive automation authority [7]. One way to adjust the level of adaptive authority is to address environmental factors, including vehicle position, yaw rate, etc. [10, 11]. Another way is to adjust the authority based on the driver states, which has been drawn little attention.
Germany, Karolinenplatz 5, 64289 Darmstadt, Germany (e-mail: [email protected]) E. Nacpil is with the Institute of Industrial Science, The University of Tokyo, Tokyo 153-8505, Japan (e-mail: [email protected]) Z. Yan is with the Institute of Industrial Science, The University of Tokyo, Tokyo 153-8505, Japan (e-mail: [email protected]) K. Nakano is with the Institute of Industrial Science, The University of Tokyo, Tokyo 153-8505, Japan (e-mail: [email protected]).
Adaptive driver-automation shared steering control via forearm surface electromyography measurement
Zheng Wang , Member, IEEE,
Satoshi Suga, Edric John Cruz Nacpil, Zhanhong Yan, and Kimihiko Nakano,
Member, IEEE D II. E XPERIMENTAL M ETHOD A. Participants
Ten healthy male subjects were recruited to participate in the experiment. Their age ranged from 20 to 25 (mean = 22, SD = 1.8). All had a valid Japanese driver’s license with a driving experience (mean = 2.6 years, SD = 1.3). In response to the question of driving frequency, two participants reported driving once a week, and others less than once a week. They all had normal or corrected-to-normal vision when performing the driving tasks in the experiment. The experiment was approved by the Office for Life Science Research Ethics and Safety, Graduate School of Interdisciplinary Information Studies, the University of Tokyo (No. 14 in 2017). Each participant received monetary compensation for the involvement in the experiment. B. Apparatus
As shown in Fig. 2, the experiment was conducted in a moving-based driving simulator with brake and accelerator pedals, an actuated steering wheel, and an instrument dashboard. The driving simulator is considered to be high-fidelity because it includes a 140-degree field-of-view and a moving platform with six degrees of freedom. To emulate the feeling of on-road driving, high frequency vibrations were produced by the moving platform, engine sounds were provided by two stereo speakers, and a self-aligning torque was generated by the actuated steering wheel. Raw data of driving performance were recorded in the host computer of the driving simulator at a sample rate of 120 Hz. A Myo armband system (Thalmic Labs, Inc.) was employed to measure the sEMG signal of driver’s dominant forearm. Compared to other commercially available armbands, Myo armband provides software development kit which allows developers to obtain access to measured data. It has been widely used in research experiments and applications [22, 23]. The muscle activity was measured by calculating the root mean square (RMS) value of the activation from eight sEMG sensors within the armband [24]. Normalization of driver grip strength was performed by measuring the maximum sEMG value for
Fig. 1. Schematic diagram of shared steering control. Fig. 2. High-fidelity driving simulator used in the experiment. C. Haptic Guidance Steering System
In the driving simulator, an electronic steering system was connected to the host computer through a CAN (controller area network) communication. The electronic steering system mainly consisted of a steering wheel, a servo motor and an electronic control unit (ECU), as shown in Fig. 3. The variables shown in Fig. 3 are defined in Table I. After calculating the real-time haptic guidance torque, the host computer inputted it to the ECU to actuate the servomotor that applied haptic guidance torque to the steering wheel. The maximum motor torque was 5 N·m, and the motor reduction ratio was 1/14. Accordingly, the steering wheel system can provide 70 N·m as a maximum active torque. The sensor resolution was 0.1° for the steering wheel angle, and 0.005 N·m for the steering torque. The magnitude and direction of haptic guidance torque were determined by comparing the vehicle trajectory and target trajectory. It is important to notice that the system is designed to assist the driver, instead of controlling the vehicle automatically. Consequently, the driver can choose to overrule the system at any time by exerting more torque on the steering wheel. In this experiment, the haptic guidance torque was limited to 5 N·m [6]. Parameters of the vehicle trajectory and target trajectory are presented in Fig. 4. The target trajectory is generated by a 5 th degree Bezier curve that was previously reported to achieve smooth lane changing performance [25, 26]. The coordinates of the target trajectory were stored in the host computer of the driving simulator. The haptic guidance torque was calculated based on a model with two look-ahead points [27]. The lateral error at the near point, 𝑒 !( , is defined as the distance between the position of vehicle and target trajectory at the near point, and the yaw error at the far point, 𝑒 (()%&) , is defined as the angle between the direction of vehicle and target trajectory at the far point. The look-ahead time of the near point is 0.3 s, and the far point is 0.7 s. The haptic guidance torque, T h , is expressed as 𝑇 * = 𝐾(𝑎 + 𝑒 !( + 𝑎 , 𝑒 (()%&) ) (1) where 𝑎 + and 𝑎 , are constant gains for 𝑒 !(-./0) , and 𝑒 ((1/0) , respectively; K is constant gain for the haptic guidance torque. The gains of 𝑎 + and 𝑎 , were determined as 0.19 and 3.8 respectively by a trial-and-error process. The settings of shared steering control authority were inspired by [12], in which different types of fixed and adaptive haptic guidance were addressed. In this experiment, different types of haptic guidance were determined by the value of K , as shown in Table II: The haptic guidance with a strong feedback gain (HG-Strong, K = 1.0) was half of the gain set for automated double lane change; the haptic guidance with a normal feedback gain (HG-Normal, K = 0.5) was half of the gain set for HG-Strong; in the condition of HG-Decrease, the gain of haptic guidance decreases linearly from 1 to 0, when the grip strength increases from 0 to the maximum value, sEMG REF ; and in the condition of HG-Increase, the gain of haptic guidance increases linearly from 0 to 1, when the grip strength increases from 0 to the maximum value, sEMG
REF . Moreover, if the grip strength exceeded sEMG
REF during the double lane change task, the gain of haptic guidance gain was set to 0 for HG-Decrease, and 1.0 for HG-Increase.
Fig. 3. Diagram of steering control loop. TABLE
I S
YMBOLS IN THE S TEERING C ONTROL L OOP
Definition 𝑒 !( Lateral error at the near point 𝑒 (()%&) Yaw error at the far point I Motor control current T d Driver input torque T h Haptic guidance torque T a Aligning torque 𝜑 Steering wheel angle
Fig. 4. Diagram of two-point haptic guidance model following target trajectory. TABLE
II E
XPERIMENTAL C ONDITIONS WITH D IFFERENT T YPES OF H APTIC G UIDANCE
Condition Description Feedback gain ( K ) Manual No haptic guidance 0 HG-Strong Fix authority with strong feedback 1.0 HG-Normal Fix authority with normal feedback 0.5 HG-Decrease Adaptive authority with decreased feedback as grip strength increases *+,
HG-Increase Adaptive authority with increased feedback as grip strength increases 𝑠𝐸𝑀𝐺𝑠𝐸𝑀𝐺 *+, D. Experimental Conditions and Scenario
The participants drove under five conditions, as shown in Table II, each with a different type of haptic guidance: (1) No haptic guidance (Manual), (2) haptic guidance with a strong feedback gain (HG-Strong), (3) haptic guidance with a normal feedback gain (HG-Normal), (4) haptic guidance decreases when grip strength increases (HG-Decrease), and (5) haptic guidance increases when grip strength increases (HG-Increase). Two Latin squares were used to partially counterbalance the within-subject order of the conditions among the 10 participants. The first Latin square with all five experimental conditions was used for the first through fifth participant, whereas the second Latin square for the sixth through tenth participant mirrored first Latin Square. As shown in Fig. 5, the driving environment was a two-lane expressway with lanes 3.6 m wide and an emergency lane on the left. Lane markings were solid lines and dashed lines. An overhead view of the double lane change track with cones is shown in Fig. 6. The driving speed of the ego vehicle was fixed at 50 km/h by a PID controller; thus, the participants did not need to operate the accelerator and brake pedals. Since steering and speed are interdependent, the speed was fixed to prevent it from confounding the assessment of steering performance. E. Procedure
Prior to participating in the experiment, test subjects signed a consent form explaining the procedure of the experiment. Each participant mounted the Myo armband on the dominant forearm. Experiment preparation consisted of two procedures: (1) Calibration and (2) normalization. In order to calibrate the Myo armband for each participant, the default program supplied by the armband manufacturer was used. Normalization of grip strength was realized by having each participant grip the driving simulator steering wheel in a “ten-to-two” position for two seconds with maximum effort. This procedure was repeated three times with 10 s of rest between each repetition. The mean value across all repetitions for a given participant was used as the reference sEMG value, sEMG
REF , for normalization. After the preparation, the participants boarded the driving simulator, and were instructed to grab the steering wheel in the “ten-to-two” position and to follow Japanese traffic regulations. Before the experimental session, the participants were required to perform a practice task to become familiar with the driving simulator. In the formal experimental session, the participants performed the double lane change task five times under each of the five driving conditions. Therefore, there were 25 trials of double lane change per participant. After each condition, the participants were asked to complete a questionnaire to assess subjective task load. The entire experiment took approximately 90 min per participant. F. Measures
The double lane change performance was measured by driver steering behavior, lane departure risk, sEMG signal, and subjective assessment on driver workload. Driver steering behavior was evaluated by measuring driver input torque and steering wheel angle. RMS value of driver input torque was calculated to evaluate driver steering effort. The steering contro l activity was evaluated by calculating the RMS value of steering wheel angle and steering wheel peak angle. The steering wheel peak angle includes maximum positive steering wheel angle and minimum negative steering wheel angle. Lane departure risk during the double lane change was evaluated by calculating lateral error with respect to the centerline of the lane. The double lane change process consists of two parts: first lane change part and second lane change part. At the first lane change ending point when the vehicle drives parallel with respect to the entered lane, the lateral error to centerline of lane was calculated. At the second lane change ending point, the driver normally made a small adjustment to get back to the centerline of lane which is called overshoot, and the lateral error to centerline of lane was calculated. The sEMG signals of forearm muscle were measured to examine driver grip strength during the double lane change task. The RMS value of normalized sEMG signal (s
EMG /s EMG
REF ) was calculated, and a higher value would indicate a larger grip strength. The perceived workload of driving task was rated using the NASA-TLX [28]. The participants were asked to use the NASA-TLX to assess their workload at the end of each driving condition. In the first step, each item of the index was investigated separately to obtain the scale score. In the second step, an individual weighting of the items was explored to obtain the weighted score. The weighted score was multiplied by the scale score for each item, and then the overall task load score was obtained. G. Data Analysis
Data were statistically analyzed using one-way repeated measures ANOVA with the Fisher-Hayter Post Hoc test to determine whether there was any significant difference between
Fig. 5. Driving environment in the experiment. Fig. 6. Illustration of double lane change track with cones. p < 0.05 to reject null hypothesis that there was no significant difference. When a p -value is between 0.05 and 0.1, it is interpreted as a tendency towards significant difference. III. R ESULTS
In this section, results are presented separately for driver steering behavior, lane departure risk, sEMG measurement and subjective assessment of driver workload. The results indicate that driver behavior was significantly influenced by different types of haptic guidance. Table III lists the means and standard deviations, results of one-way repeated measures ANOVA, and pairwise comparisons based on the measurements of driver behavior. A. Driver Steering Behavior
The results of driver steering behavior were obtained by measuring driver input torque and steering wheel angle. As indicated in Table III, the RMS of driver input torque was significantly different among the five conditions. Fig. 7 shows the result of RMS of driver input torque with error bar among 10 participants. Driver input torque was significantly higher in the condition of Manual than other conditions, which indicates that haptic guidance significantly reduced driver steering effort. HG-Strong induced significantly lower driver input torque compared to other conditions, and HG-Decrease was in between of HG-Strong and HG-Normal. Moreover, HG-Decrease induced lower driver input torque compared to HG-Increase. As indicated in Table III, RMS of steering wheel angle was significantly different among the five conditions. Fig. 8 shows the results of RMS of steering wheel angle with error bars among 10 participants. It can be observed that HG-Decrease yielded lower steering wheel angle compared to HG-Normal and HG-Strong, and Manual yielded lower steering wheel angle compared to HG-Strong. Therefore, driver steering control activity was to some extent reduced by HG-Decrease and increased by HG-Strong. The result of steering wheel angle is also indicated by plotting the averaged steering wheel angle with standard
TABLE
III M
EANS AND S TANDARD D EVIATIONS OF THE D EPENDENT M EASURES OF D RIVER B EHAVIOR
Variable Manual (1) M(SD) HG- Strong (2) M(SD) HG- Normal (3) M(SD) HG- Dec (4) M(SD) HG- Inc (5) M(SD) p RMS of driver input torque (N·m) *** *** *** *** *** 0.49 *** * 0.12 *
RMS of SWA (deg) + + ** Maximum positive value of SWA (deg) *** 0.66 0.76 0.98 0.27 ** 0.11 0.19 0.89 0.20
Minimum negative value of SWA (deg) -44.5 (14.4) -47.2 (16.9) -44.1 (12.8) -41.0 (12.7) -44.4 (12.3) 0.032
Lateral error at the end of 1st LC (m) + * + * Lateral error at the end of 2nd LC (m) * RMS of normalized sEMG (%) + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001 SWA: Steering wheel angle; LC: Lane change. Fig. 7. RMS of driver input torque. Data error bars represent mean +/- SEM (standard error of mean). Manual HG-Strong HG-Normal HG-Decrease HG-Increase
Driving conditions R M S o f d r i v er i npu t t o r qu e ( N · m ) SEMS1S2S3 S4S5S6S7 S8S9S10
Fig. 8. RMS of steering wheel angle. Data error bars represent mean +/- SEM.
Manual HG-Strong HG-Normal HG-Decrease HG-Increase
Driving conditions R M S o f s t eer i n g w h ee l a n g l e ( d e g ) SEMS1S2S3 S4S5S6S7 S8S9S10 B. Lane Departure Risk
The lane departure risk during the double lane change was evaluated by respectively calculating the lateral error from centerline of lane at the ending points of the first lane change part and second lane change part. As indicated in Table III, lateral error at the end of first lane change part was significantly different among the five conditions. Fig. 10 shows the lateral error from centerline of lane with error bars among 10 participants. The lateral error was lower for HG-Decrease than for Manual, HG-Normal and HG-Increase, and lower for HG-Strong than for Manual. Therefore, HG-Decrease and HG-Strong reduced the lane departure risk compared to Manual. As indicated in Table III, lateral error at the end of second lane change part was not significantly different among the five conditions. According to pairwise comparisons, the lateral error in the condition of HG-Decrease was lower than of HG-Increase. Fig. 11 shows the overshoot from centerline of lane with error bars among 10 participants. There is a tendency that HG-Decrease yielded a relatively lower lane departure risk compared to other conditions. C. sEMG Measurement As indicated in Table III, RMS of normalized sEMG was not significantly different among the five conditions. Fig. 12 shows the results of RMS of normalized sEMG with error bars among 10 participants. There is a tendency that participants used more
Fig. 10. Lateral error from centerline of lane at the end of first lane change part. Data error bars represent mean +/- SEM.
Manual HG-Strong HG-Normal HG-Decrease HG-Increase
Driving conditions L a t er a l err o r f r o m t h e ce n t er li n e o f l a n e ( m ) SEMS1S2S3 S4S5S6S7 S8S9S10
Fig. 11. Overshoot (lateral error) from centerline of lane at the end of second lane change part. Data error bars represent mean +/- SEM.
Manual HG-Strong HG-Normal HG-Decrease HG-Increase
Driving conditions O v er s h oo t f r o m ce n t er li n e o f l a n e ( m ) SEMS1S2S3 S4S5S6S7 S8S9S10
Fig. 9. Averaged steering wheel angle with standard deviation among five double lane change of all participants.
200 220 240 260 280 300 320
Position X (m) -60-40-200204060 S t eer i n g w h ee l a n g l e ( d e g ) ManualHG-StrongHG-NormalHG-DecreaseHG-Increase
Fig. 12. RMS of normalized sEMG. Data error bars represent mean +/- SEM.
Manual HG-Strong HG-Normal HG-Decrease HG-Increase
Driving conditions R M S o f n o r m a li ze d s E M G ( % ) SEMS1S2S3 S4S5S6S7 S8S9S10 D. Subjective Assessment of Driver Workload
Subjective assessment of driver workload was rated by using NASA-TLX, and the results with error bars are shown in Fig. 13. Based on One-way repeated measures ANOVA, the statistical analysis shows that there was significant difference in Effort among the five conditions, where p = 0.042, and no significant difference for other parameters (Mental Demand: p = 0.215, Physical Demand: p = 0.236, Temporal Demand: p = 0.341, Performance: p = 0.799, Frustration: p = 0.154, Overall Workload: p = 0.220). The pairwise comparison results show that Effort was significantly lower for HG-Decrease than for Manual ( p < 0.05), and significantly lower for HG-Increase than for Manual ( p < 0.01). In addition, HG-Decrease ( p = 0.12) and HG-Increase ( p = 0.12) tended to reduce the overall workload compared to Manual, although the difference was not significant. As shown in Fig. 13, there is a tendency that the driver workload could be reduced by haptic guidance compared to manual driving. As for the condition of HG-Decrease, the mean value shown in the figure is lower compared to the other conditions in terms of Mental Demand, Physical Demand, Effort, Frustration, and Overall Workload. Thus, HG-Decrease is to some extent more effective at reducing driver workload. IV. D ISCUSSION
Different types of driver–automation shared control were realized by using a haptic guidance steering system that was configured for fixed authority (HG-Strong and HG-Normal) and adaptive authority (HG-Decrease and HG-Increase). By looking at the results of driver steering behavior, both driver input torque and steering wheel angle were relatively lower for the condition of HG-Decrease, indicating that HG-Decrease yielded a lower steering control effort and control activity. This result is in accordance with the result of subjective assessment on driver workload by NASA-TLX, in which the driver workload was found to be lower for the condition of HG-Decrease. It is interesting that HG-Strong significantly reduced the driver input torque but increased the steering wheel angle. In some of the previous studies, only one of these two evaluation indexes was used to assess driver steering workload [8, 27]. HG-Strong shows controversial results in terms of these two indexes, indicating that both indexes should be addressed when analyzing driver steering workload. It was expected that haptic guidance system would reduce lane departure risk relative to manual driving, and HG-Decrease would perform better relative to other haptic guidance conditions. The idea of designing HG-Decrease is to reduce the authority of haptic guidance when the driver increases the grip strength to gain more control authority. Although it was observed that HG-Decrease yielded a relatively lower lane departure risk compared to other conditions, other haptic guidance conditions did not have significantly better performance compared to manual driving. One explanation could be that the driver’s individual driving trajectory [5] could be significantly different from the optimal 5 th degree Bezier curve that was used as target trajectory in the experiment. Another explanation could be that the driving speed of the ego vehicle was fixed at 50 km/h, whereas drivers would choose their own driving speed during regular driving [17]. If the driver could drive in a more regular way, and the target trajectory could be individualized, it is expected that the haptic guidance system could significantly reduce the lane departure risk. One of the experimental hypotheses is that driver grip strength would be different among the driving conditions due to mutual adaptation between the driver and haptic guidance system. Human beings receive and integrate multiple sensory information from their surroundings, in order to understand and interact with the world. When driving with the haptic guidance system, drivers tended to integrate visual and haptic information to achieve better driving performance [8, 27]. It has been found that the driver could choose to rely on or to resist the haptic guidance by adjusting their arm admittance [3]. However, according to the sEMG measurement results, this expected outcome was not observed. One explanation could be that more time is needed for the driver to adapt to the haptic guidance system [27]. A longer driving scenario with curved roads will be designed as future work, and the mutual adaption behavior between driver and haptic guidance system will be investigated for lane change and lane following tasks. Although the current experiment was conducted in a high-fidelity driving simulator, future work will address how driver behavior would be different in a real-vehicle experiment, especially for driver workload. Moreover, the subjective preference for different types of haptic guidance systems and whether the systems are preferable under realistic driving conditions could be tested in a future experiment with a real vehicle. Another limitation of the current experiment is that using the Myo armband instead of conventional sEMG equipment may limit the applicability of the study to other sEMG-based driving simulator experiments. The comparison between conventional and commercial sEMG equipment should be addressed by an experimental study in the future. V. C ONCLUSION
This paper focuses on designing adaptive shared steering control via forearm sEMG measurement and evaluating its
Fig. 13. Mean scores on NASA-TLX (*p < 0.05 and **p < 0.01). Data error bars represent mean + SEM. A CKNOWLEDGMENT
The authors would like to thank Pan Xiuxi of Tokyo Institute of Technology for his assistance in realizing network communication between the Myo armband and haptic guidance system in driving simulator. R EFERENCES [1]
W. Wen, Y. Kuroki, and H. Asama, “The sense of agency in driving automation,”
Frontiers in Psychology , vol. 10, p. 2691, Dec. 2019. doi:10.3389/fpsyg.2019. [2]
H. Muslim and M. Itoh, “A theoretical framework for designing human-centered automotive automation systems,”
Cognition, Technology & Work , vol. 21, no. 4, pp. 685-697, Nov. 2019. [3]
D. A. Abbink, M. Mulder, and E. R. Boer, “Haptic shared control: Smoothly shifting control authority,”
Cognition, Technology & Work , vol. 14, no. 1, pp. 19-28, Mar. 2012. [4]
F. Flemisch, D. A. Abbink, M. Itoh, M. P. Pacaux-Lemoine, and G. Weßel, “Joining the blunt and the pointy end of the spear: Towards a common framework of joint action, human–machine cooperation, cooperative guidance and control, shared, traded and supervisory control,”
Cognition, Technology & Work , vol. 21, no. 4, pp. 555-568, Nov. 2019. [5]
M. Mulder, D. A. Abbink, and E. R. Boer, “Sharing control with haptics: Seamless driver support from manual to automatic control,”
Human Factors , vol. 54, no. 5, pp. 786-798, Oct. 2012. [6]
Z. Wang, R. Zheng, T. Kaizuka, K. Shimono, and K. Nakano, “The effect of a haptic guidance steering system on fatigue-related driver behavior,”
IEEE Transactions on Human Machine Systems , vol. 47, no. 5, pp. 741-748, Oct. 2017. [7]
F. Mars, M. Deroo, and J. M. Hoc, “Analysis human-machine cooperation when driving with different degrees of haptic shared control,”
IEEE Transactions on Haptics , vol. 7, no. 3, pp. 324-333, Jul.-Sep. 2014. [8]
Z. Wang, R. Zheng, T. Kaizuka, and K. Nakano, “Influence of haptic guidance on driving behavior under degraded visual feedback conditions,”
IET Intelligent Transport Systems , vol. 12, no. 6, pp. 454-462, Aug. 2018. [9]
K. Iwano, P. Raksincharoensak, and M. Nagai, “A study on shared control between the driver and an active steering control system in emergency obstacle avoidance situations,”
IFAC Proceedings Volumes , vol. 47, no.3, pp. 6338-6343, Aug. 2014. [10]
A. T. Nguyen, C. Sentouh, and J. C. Popieul, “Sensor reduction for driver-automation shared steering control via an adaptive authority allocation strategy,”
IEEE/ASME Transactions on Mechatronics , vol. 23, no. 1, pp. 5-16, Feb. 2018. [11]
Z. Wang, Z. Yan, and K. Nakano, “Comfort-oriented haptic guidance steering via deep reinforcement learning for individualized lane keeping assist,” in
Proceedings of IEEE International Conference on Systems, Man, and Cybernetics , 2019, pp. 4290-4296. [12]
J. Smisek, W. Mugge, J. B. Smeets, M. M. Van Paassen, and A. Schiele, “Adapting haptic guidance authority based on user grip,” in
Proceedings of IEEE International Conference on Systems, Man, and Cybernetics , 2014, pp. 1516-1521. [13]
J. Vogel, C. Castellini, and P. van der Smagt. “EMG-based teleoperation and manipulation with the DLR LWR-III,” In
Proceedings of IEEE/RSJ International Conference on Intelligent Robots and Systems , 2011, pp. 672-678. [14]
L. Peternel, T. Noda, T. Petrič, A. Ude, J. Morimoto, and J. Babič, “Adaptive control of exoskeleton robots for periodic assistive behaviours based on EMG feedback minimization,”
PloS One , vol. 11, no. 2, p. e0148942, Feb. 2016. [15]
C. Wong, Z. Zhang, B. Lo, and G. Z. Yang, “Wearable sensing for solid biomechanics: A review,”
IEEE Sensors Journal , vol. 15, no. 5, pp. 2747-2760, May 2015. [16]
Y. Zhou, Y. Fang, K. Gui, K. Li, D. Zhang, and H. Liu, “sEMG bias-driven functional electrical stimulation system for upper-limb stroke rehabilitation,”
IEEE Sensors Journal , vol. 18, no. 16, pp. 6812-6821, Aug. 2018. [17]
F. Mars and P. Chevrel, “Modelling human control of steering for the design of advanced driver assistance systems,”
Annual Reviews in Control , vol. 44, pp.292-302, 2017. [18]
V. Balasubramanian and R. Bhardwaj, “Grip and electrophysiological sensor-based estimation of muscle fatigue while holding steering wheel in different positions,”
IEEE Sensors Journal , vol.19, no. 5, pp. 1951-1960, Mar. 2018. [19]
H. Nakamura, D. Abbink, and M. Mulder, “Is grip strength related to neuromuscular admittance during steering wheel control,” in
Proceedings of IEEE International Conference on System, Man, and Cybernetics , 2011, pp. 1658-1663. [20]
Z. Wang, R. Zheng, T. Kaizuka, K. Shimono, and K. Nakano, “Evaluation of driver steering performance with haptic guidance under passive fatigued situation,” in
Proceedings of IEEE International Conference on System, Man, and Cybernetics , 2016, pp. 3334-3339. [21]
K. Wang, X. Zhang, J. Ota, and Y. Huang, “Development of an sEMG-handgrip force model based on cross model selection,”
IEEE Sensors Journal , vol. 19, no. 5, pp. 1829-1838, Mar. 2019. [22]
S. Pizzolato, L. Tagliapietra, M. Cognolato, M. Reggiani, H. Muller, and M. Atzori, “Comparison of six electromyography acquisition setups on hand movement classification tasks”,
Plos One , vol. 12, no. 10, p. e0186132, Oct. 2017. [23]
P. Visconti, F. Gaetani, G. A. Zappatore, and P. Primiceri, “Technical features and functionalities of Myo armband: An overview on related literature and advanced applications of myoelectric armbands mainly focused on arm prostheses,”
International Journal on Smart Sensing and Intelligent Systems , vol. 11, no. 1, pp. 1-25, 2018. [24]
L. Han, H. Yashiro, H. T. N. Nejad, Q. H. Do, and S. Mita, “Bezier curve based path planning for autonomous vehicle in urban environment,” in
Proceedings of IEEE Intelligent Vehicle Symposium , 2010, pp. 1036-1042. [26]
I. Bae, J. Moon, H. Park, J. H. Kim and S. Kim, “Path generation and tracking based on a Bezier curve for a steering rate controller of autonomous vehicles,” in
Proceedings of IEEE International Conference on Intelligent Transportation Systems , 2013, pp. 436-441. [27]
Z. Wang, R. Zheng, T. Kaizuka, and K. Nakano, “Relationship between gaze behavior and steering performance for driver-automation shared control: A driving simulator study,”
IEEE Transactions on Intelligent Vehicles , vol. 4, no. 1, pp. 154-166, Mar. 2019. [28]
S. G. Hart and L. E. Staveland, “Development of NASA-TLX (Task Load Index): Results of empirical and theoretical research,”
Advances in Psychology , vol. 52, pp. 139-183, 1988., vol. 52, pp. 139-183, 1988.