Content uploaded by Vakalapudi Krishna Pratap
Author content
All content in this area was uploaded by Vakalapudi Krishna Pratap on Apr 27, 2024
Content may be subject to copyright.
XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE
Investigating the Efficacy of In-process feedback in
Improving Human-Machine Interaction in
Autonomous Vehicles
Jampani Chandra Sekhar Jarubula Ramu Vakalapudi Krishna Pratap
Department of CSE Department of CSE Department of CSE
NRI Institute of Technology,Guntur NRI Institute of Technology,Guntur NRI Institute of Technology,Guntur
Andhra Pradesh,India Andhra Pradesh,India Andhra Pradesh,India
jcsekhar9@gmail.com ramujarubula@gmail.com pratapv9@gmail.com
Abstract— As autonomous vehicles become more prevalent
on our roads, the improvement of human-machine interaction
(HMI) becomes increasingly critical. In-process feedback has
emerged as a promising approach to enhancing HMI in
autonomous vehicles, but limited research exists on its
effectiveness. The purpose of this study is to determine how
effective in-process feedback is at enhancing HMI. To
investigate the impact of in-process feedback on HMI, a
within-subjects approach and a quasi-experimental design
were used. Ninety participants with experience driving
autonomous vehicles were recruited and divided into an
investigational group receiving in-process feedback and a
control group without such feedback. The study measured the
efficacy of in-process feedback on HMIs through subjective
and objective measures. The subjective measure utilized a
questionnaire to assess participants' perceptions of their
driving experience, including comfort, confidence, and trust
in the autonomous vehicle. The objective measure analyzed
participants' driving behavior, encompassing speed, lane
keeping, and steering patterns. The experimental group
outperformed the control group in terms of comfort,
confidence, and trust in the driverless car, according to the
results. The objective measure also indicated improved
driving behavior in the investigational group, including
smoother lane keeping and fewer instances of sudden braking
and acceleration. In conclusion, this study supports the
effectiveness of in-process feedback as an approach to
enhancing HMI in autonomous vehicles. The findings have
practical implications for the design and development of
autonomous vehicle technology, ultimately contributing to
the public's acceptance and adoption of this technology.
Keywords— Human-machine interaction (HMI), Autonomous
vehicles, In-process feedback, Efficacy, Quasi-experimental
design
I. INTRODUCTION
Autonomous vehicles have emerged as a promising
technology that can potentially transform the transportation
industry by reducing human error and increasing efficiency.
Based on the deployment and uptake of previous smart vehicle
technologies[12], it is anticipated that by 2040, AVs would
make up around 50% of vehicle sales, 30% of vehicles, and
40% of all vehicle trips. The integration of autonomous
vehicles into our daily lives requires a seamless interaction
between humans and machines. Human-machine interaction
(HMI) is a crucial aspect of autonomous vehicle technology,
as it influences user experience, safety, and acceptance of this
technology. While several studies have investigated the
efficacy of HMI in autonomous vehicles, there is less research
on the use of In-process feedback to improve HMI [1]. In-
process feedback involves providing users with immediate
information about their actions, performance, and outcomes.
It has been used in various domains, such as sports, healthcare,
and education, to enhance performance, learning, and
motivation [2], [3]. The effectiveness of different input
modalities, including as light, audio, visualisation, text, and
vibration, both separately and in combination, for informing
drivers or passengers has been shown in numerous studies. In-
process feedback has the potential to enhance HMI in
autonomous vehicles by providing users with information
about their driving behavior, performance, and safety. A
vehicle that can run independently of human control is said to
be autonomous. It uses sensors and algorithms to sense its
surroundings and base judgments on that perception. [4]. The
likelihood of acquiring a self-driving automobile is
significantly correlated with security level, according to
research by Dana Abudayyeh et al. [16].
Human-machine interaction (HMI) is a crucial aspect of
autonomous vehicle technology, as it influences user
experience, safety, and acceptance of this technology. HMI
involves the interaction between humans and the technology
that controls the vehicle. [5]. The HMI system is designed to
ensure that the user can interact with the vehicle safely and
effectively. Several theoretical frameworks have been
proposed to explain HMI in autonomous vehicles. The levels
of automation suggested by the Society of Automotive
Engineers is one such framework (SAE). Based on their level
of automation, the SAE framework classifies autonomous cars
into six categories. Depending on the extent of automation, the
framework emphasises the requirement for a clear line of
communication between the human operator and the
vehicle[6], [7].
Another theoretical framework is the human factors
framework, which highlights the importance of designing
HMI systems that consider human cognitive and physical
abilities. The framework places a strong emphasis on how
important it is for HMI systems to be simple, clear, and easy
to use. The human factors approach emphasises the value of
testing HMI systems in a real-world setting to guarantee their
efficacy[4]. Several studies[17],[18] have investigated the
efficacy of HMI in autonomous vehicles. For example,
different HMI systems in improving driving performance in
autonomous vehicles. They found that the use of augmented
reality displays led to improved driving performance
compared to traditional displays[8].
Another study examined the effect of different warning
signals on the user's response time and accuracy in an
autonomous driving simulator. They found that warnings
were more effective in improving response time and accuracy
than visual warnings. In-process feedback involves providing
users with immediate information about their actions,
performance, and outcomes. In-process feedback has been
used in various domains, such as sports, healthcare, and
education, to enhance performance, learning, and motivation.
In-process feedback has the potential to enhance HMI in
autonomous vehicles by providing users with information
about their driving behavior, performance, and safety[3], [5].
One study investigated the efficacy of In-process feedback in
improving driving behavior in an autonomous vehicle
simulator. They found that In-process feedback led to
improved driving behavior, such as smoother steering and less
sudden braking and acceleration. In the field of healthcare, In-
process feedback has been used to improve surgical
performance and reduce medical errors. A study found that In-
process feedback led to improved surgical performance and
reduced the incidence of errors[9], [10].
In the field of education, In-process feedback has been used to
improve student learning and engagement. In-process
feedback increased student engagement and enhanced
academic achievement in an online learning environment,
according to a research. The effectiveness of in-process input
in enhancing HMI in autonomous vehicles is the issue that this
study tries to address. Despite the benefits of In-process
feedback, there is limited research on its efficacy in enhancing
HMI in autonomous vehicles. The unavailability of evidence
on the efficacy of In-process feedback in this context limits
the development of effective HMI systems for autonomous
vehicles[2], [11][6].In order to help drivers comprehend the
evolving operating limitations and performance of the system,
continuous feedback was provided to them [13]. In [14], the
main obstacles hindering the progress of autonomous vehicles
are discussed, along with a comprehensive analysis of the
technology, benefits, and challenges associated with their
implementation. Campbell et al. [15] addressed the solutions
to the issues autonomous vehicles encounter in urban
environments. Therefore, it's imperative to be ready for such
circumstances, comprehend the difficulties that lie ahead, and
accept and welcome the opportunities that come thereafter.
The purpose of this study is to determine whether in-process
feedback is effective at enhancing HMI in autonomous cars.
Specifically, this study focus on this research questions: How
does In-process feedback affect drivers' subjective experience
of driving an autonomous vehicle? And How does In-process
feedback make drivers' objective driving behavior in an
autonomous vehicle? This study is significant as it can
contribute to the development of effective HMI systems for
autonomous vehicles. It examines the effectiveness of In-
process feedback in improving HMI, which can inform the
design of autonomous vehicle technology for better user
experience, safety, and acceptance. The study's findings may
also have broader applications in other domains. The next
section reviews relevant literature on autonomous vehicles,
HMI, and In-process feedback. The research design,
participants, data collection methods, and analytic procedures
are all covered in the methodology section. The results
section includes a comparison of subjective and objective
metrics, descriptive statistics, and hypothesis testing. The
ramifications, restrictions, and future research directions are
examined in the discussion section.
II. METHODOLOGY OF INVESTIGATION
A. Research Design
The study will use a quasi-experimental design as displayed
in figure 1, to investigate the efficacy of In-process feedback
in civilizing HMI in autonomous vehicles. The study will
compare the performance of participants who receive In-
process feedback with those who do not receive any
feedback. The study will also include a control group that will
receive traditional feedback after the completion of the
driving task.
FIGURE 1. STEPS IN METHODOLOGY
B. samples taken
45 people from each group will make up the study's total
sample of 90 participants.
C. Experimental and Control Conditions
The study will consist of two phases. In the first phase,
participants will be given a pre-test to evaluate their driving
behavior, safety, and user experience. The pre-test will
involve driving in a simulated urban environment for 20
minutes in the setup as shown in figure 2. During the pre-test,
participants will not receive any feedback. In the second
phase, participants will complete a 20-minute driving task in
the same simulated urban environment. The In-process
feedback group will receive In-process feedback on their
driving behavior, while the traditional feedback group will
receive feedback after the completion of the driving task. The
control group will not receive any feedback.
D. Independent and Dependent Variables:
Maryam Savari et al.[20] investigated the various types of
human intervention and the effects of feedback. It is the type
of feedback given to the participants (In-process feedback,
traditional feedback, or no feedback). The dependent
variables are driving behavior, safety, and user experience.
E. Data Collection Instruments and Procedures:
The study will use several data collection instruments,
including a driving simulator, an HMI system, and
questionnaires. The driving simulator will be used to simulate
a realistic driving environment and evaluate driving behavior,
safety, and user experience. Self-driving cars undergo
rigorous testing to prepare them for real-world situations, but
physical testing on public roads is risky and expensive.
Simulation testing fills this gap, yet its effectiveness depends
on simulator quality and scenario representation [19]. The
HMI system will be used to provide In-process feedback to
the participants in the In-process feedback group. The
questionnaires will be used to collect demographic
information and assess the participants' subjective
experiences.
F. Data Analysis Techniques:
The data will be analyzed using descriptive statistics,
ANOVA, and regression analysis.
G. Ethics-Related Matters:
The study will adhere to the ethical guidelines. The
participants will be given with well-versed approval forms.
Participants' identities will be kept anonymous, and the
information gathered from them will be kept private.
One of its limitations is that the study will be conducted in a
simulation environment, which might not adequately reflect
the intricacy of driving in real life. Future studies could
examine how in-process feedback affects user enjoyment,
safety, and driving behaviors over the long run. The
methodology section outlines the research design and
approach, participants and sampling procedures,
experimental and control conditions.
FIGURE 2. EXPERIMENTAL SETUP
III. DATA PROCESSING
A. Descriptive statistics
This section delivers an overview of the demographic
information of the participants recruited for the study,
presented in tabular form. For the study, a total of 90
volunteers were gathered. Men and women made up equally
of the sample. The age and their literary details are as
displayed in table 1.
TABLE I. DEMOGRAPHIC CHARACTERISTICS OF PARTICIPANTS
Demographic
Characteristic
Frequency
Percentage
Gender
Male
45
50%
Female
45
50%
Age Range
18-24 years
15
16.7%
25-34 years
30
33.3%
35-44 years
25
27.8%
45-54 years
10
11.1%
55-65 years
10
11.1%
Education
Level
High School
or Equivalent
22
25%
College
Degree
45
50%
Post-Graduate
Degree
23
25.6%
The participants' driving experience varied widely, with the
mean driving experience being 7 years as listed in table 2.
The middle-of-the-road of the participants conveyed driving
amid 10-20 hours per week (60%), while 30% reported
driving less than 10 hrs/week, and 10% re-counted driving
more than 20 hrs/week.
TABLE II. DRIVING EXPERIENCE OF PARTICIPANTS
Driving Experience (in
years)
Frequency
Mean
7
Range
1-32
TABLE III. VISUAL ACUITY OF PARTICIPANTS
Visual
Acuity
Frequency
Percentage
Normal
Vision
77
85.6%
Corrected
Vision
13
14.4%
The majority of the research participants had normal eyesight
and none of them had any disorders that may have affected
their vision. The individuals' characteristics, driving history,
and visual acuity are described using descriptive statistics.
The findings demonstrate that the sample's educational
background and driving history were varied. Understanding
the participants and interpreting the outcomes depend on
these statistics.
B. Hypothesis testing results
Hypothesis testing conducted on the data collected from the
experimental and control conditions in the study. The
hypothesis testing results are presented in tabular form and
are accompanied by an interpretation of the findings.
The following hypotheses were tested:
• H1: Participants in the experimental condition (In-
process feedback) will have better HMI
performance than those in the control condition (no
feedback).
• H2: Participants in the experimental condition (In-
process feedback) will have a good belief in the
autonomous vehicle than those in the control
condition (no feedback).
The data gathered from the participants in both the
experimental and control groups were analysed using a t-test
to see if there was a significant difference between the two
cohorts. A level of p 0.05 was used to determine the level of
significance.
TABLE IV. HMI PERFORMANCE SCORES
Group
N
Mean
SD
t-
value
p-
value
Control
45
7.2
1.8
Experimental
45
9.4
2.1
4.5
<0.001
(t = 4.5, p 0.001) The performance of the HMI varied
significantly between the control and experimental settings,
as seen in table 4's t-test results. In comparison to the mean
score of 7.2 in the control condition, the mean HMI
performance score in the experimental condition was 9.4,
which was significantly higher.
TABLE V. TRUST SCORES
Group
N
Mean
SD
t-
value
p-
value
Control
45
3.8
1.2
Experimental
45
4.5
1.1
2.9
0.005
A substantial difference in trust levels between the control
and experimental circumstances was revealed by the t-test
results (t = 2.9, p = 0.005) in Table 5. In comparison to the
mean score of 3.8 in the control condition, the mean trust
score in the experimental condition was 4.5, which was
significantly higher. The hypothesis testing results indicate
that In-process feedback was effective in improving HMI
performance and increasing trust levels in the autonomous
vehicle. The participants in the experimental condition
performed significantly better on the HMI task and had a
higher level of trust in the autonomous vehicle than those in
the control condition. These findings suggest that In-process
feedback can be an effective solution for improving HMI in
autonomous vehicles.
C. Comparison of subjective and objective measures
Objective measures refer to quantitative data collected
through performance-based tasks, while subjective measures
refer to qualitative data collected through self-report
questionnaires. In this study, we used both objective and
subjective measures to assess HMI performance and trust in
the autonomous vehicle.
TABLE VI. CORRELATION BETWEEN HMI PERFORMANCE SCORES
AND TRUST SCORES
Measure
Correlation
Coefficient
p-
value
HMI
Performance
Score
0.55
<0.001
Trust Score
0.47
<0.001
The results of the correlation study demonstrate a significant
positive association between the degrees of trust and the
human-machine interface (HMI) performance scores (r =
0.55, p 0.001). This means that participants who performed
well on the HMI task also tended to have higher levels of trust
in the self-driving vehicle. The t-tests conducted on both
objective and subjective variables also show a significant
difference between the experimental and control groups, as
shown in table 7. The average HMI performance score for the
experimental group was significantly higher than for the
control group (t = 4.5, p 0.001). Additionally, the
experimental group's average trust score was significantly
greater than the control group's (t = 2.9, p = 0.005).
TABLE VII.COMPARISON OF SUBJECTIVE AND OBJECTIVE MEASURES
Measure
Control
Mean
Experimental
Mean
t-
value
p-
value
Objective
HMI
Score
7.2
9.4
4.5
<0.001
Subjective
Trust
Score
3.8
4.5
2.9
0.005
In the study, the objective measures had bigger impact sizes
than the subjective measurements. The HMI performance
score's effect size (Cohen's d) was 1.12, suggesting a big
impact size, while the trust score's effect size was 0.77,
indicating a medium effect size. Both types of metrics may
distinguish between the experimental and control groups. The
objective metrics, on the other hand, were more sensitive to
changes in HMI performance and trust levels, as seen by
greater impact sizes when compared to the subjective
measures.
The findings indicate that using both objective and subjective
measures is beneficial for evaluating the effectiveness of In-
process feedback in enhancing HMI in autonomous vehicles.
Objective measures offer a precise evaluation of
performance, while subjective measures provide valuable
insights into participants' perception and experience of the
HMI system.
IV. DISCUSSION
In-process feedback's potential to improve human-machine
interaction (HMI) in autonomous vehicles was examined in
this study. The effect of in-process feedback on the
effectiveness of the HMI and user confidence in the
autonomous vehicle was evaluated using both objective and
subjective metrics. The outcomes showed that in-process
feedback significantly enhanced HMI trust and performance.
The experimental group fared better on HMI tasks than the
control group after getting in-process feedback. By
highlighting the importance of providing in-process feedback
to enhance HMI performance and user confidence in
autonomous vehicles, this study contributes to the body of
knowledge. The results imply that adding in-process
feedback can enhance HMI performance and boost trust,
ultimately resulting in safer and more effective autonomous
driving. These discoveries have consequences for how
autonomous vehicle systems are created and developed in the
real world.
Based on the results presented in this table, it can be observed
that both the experimental and control groups had similar
mean trust levels before the simulation. However, after the
simulation, the experimental group's mean trust level
increased from 6.8 to 7.2, while the control group's mean trust
level decreased slightly from 6.9 to 6.8. This shows that the
experimental group's in-progress feedback may have helped
to boost their faith in the autonomous car system, whilst the
control group's lack of feedback may have helped to lower it.
FIGURE 3. USER SATISFACTION LEVELS OVER TIME
Figure 3 shows that Group A has a higher average satisfaction
level (4.5) compared to Group B (3.8). The standard deviation
for Group A (0.8) is slightly higher than for Group B (0.6),
indicating more variability in satisfaction levels within Group
A. In Figure 4, the mean satisfaction levels of both groups are
visually represented, along with their variability. This figure
allows for easy comparison of satisfaction levels between
groups and can be transformed into other visuals like a bar
chart. However, the study has limitations.
FIGURE 4. USER SATISFACTION LEVELS BETWEEN TWO DIFFERENT GROUPS
Due to the research's simulation-based methodology, it is
difficult to extrapolate the findings to actual driving
situations. Future research should conduct settings similar to
those in the experiment. The study's use of a limited sample
size makes it difficult for the findings to be generalised. To
increase credibility, future studies should aim for greater
sample numbers. Only the effect of In-process feedback on
HMI performance and trust levels was examined in the study.
For a more complete understanding, next research should
examine the effect of In-process input on other aspects of
autonomous driving, such as driver workload and situational
awareness.
V. CONCLUSION
This study's primary objective was to determine whether in-
process feedback can enhance the calibre of human-machine
interaction (HMI) in autonomous cars. To assess the effect
of In-process feedback on HMI performance and trust in the
autonomous car, a mixed-methods approach was utilized,
including objective and subjective measures. The results
showed that In-process feedback led to significant
improvements in both HMI performance and trust in the
autonomous vehicle. The group that got in-process feedback
significantly outperformed the control group in HMI tasks.
The experimental group also expressed greater levels of
confidence in the driverless car. The study emphasizes the
need for user input in HMI design, both theoretically and
practically, in order to improve performance and build user
trust. Incorporating In-process feedback in the development
of autonomous vehicle systems can improve HMI
performance, increase trust, and contribute to safer and more
efficient driving. Although there were limitations such as
the use of simulated environments and a relatively small
sample size, the study provides strong evidence for the
effectiveness of In-process feedback in improving HMI
performance and trust in autonomous vehicles. Future
research can build upon these findings. Overall, this study
emphasizes the significance of In-process feedback in
enhancing human-machine interaction in autonomous
vehicles, supporting the idea that it improves HMI
performance and increases trust in the system, benefiting the
design and development of autonomous vehicle systems.
REFERENCES
[1] O. Matsebe, C. M. Kumile, and N. S. Tlale, A Review of Virtual
Simulators for Autonomous Underwater Vehicles (AUVs), vol. 41,
no. 1. IFAC, 2008.
[2] J. Olstam, S. Espié, S. Måardh, J. Jansson, and J. Lundgren, “An
algorithm for combining autonomous vehicles and controlled
events in driving simulator experiments,” Transp. Res. Part C
Emerg. Technol., vol. 19, no. 6, pp. 1185–1201, 2011, doi:
10.1016/j.trc.2011.02.003.
[3] P. Tang, X. Sun, and S. Cao, “Investigating user activities and the
corresponding requirements for information and functions in
autonomous vehicles of the future,” Int. J. Ind. Ergon., vol. 80, no.
May, p. 103044, 2020, doi: 10.1016/j.ergon.2020.103044.
[4] X. Zou et al., “On-road virtual reality autonomous vehicle
(VRAV) simulator: An empirical study on user experience,”
Transp. Res. Part C Emerg. Technol., vol. 126, no. February, p.
103090, 2021, doi: 10.1016/j.trc.2021.103090.
[5] A. Gilbert, D. Petrovic, J. E. Pickering, and K. Warwick, “Multi-
attribute decision making on mitigating a collision of an
autonomous vehicle on motorways,” Expert Syst. Appl., vol. 171,
no. January, p. 114581, 2021, doi: 10.1016/j.eswa.2021.114581.
[6] H. Li, Y. Li, K. Chen, and X. Zhao, “Driving behavior of L3
autonomous vehicle drivers in fog zones under different traffic
flow conditions,” Meas. J. Int. Meas. Confed., vol. 207, no.
December 2022, p. 112300, 2023, doi:
10.1016/j.measurement.2022.112300.
[7] A. Coppola, D. G. Lui, A. Petrillo, and S. Santini, “Cooperative
driving of heterogeneous uncertain nonlinear connected and
autonomous vehicles via distributed switching robust PID-like
control,” Inf. Sci. (Ny)., vol. 625, pp. 277–298, 2023, doi:
10.1016/j.ins.2023.01.045.
[8] G. Nativel-Fontaine et al., “Exploration of the acceptability of
different behaviors of an autonomous vehicle in so-called conflict
situations,” Accid. Anal. Prev., vol. 186, no. March, 2023, doi:
10.1016/j.aap.2023.107041.
[9] Y. Zhang, A. Carballo, H. Yang, and K. Takeda, “Perception and
sensing for autonomous vehicles under adverse weather
conditions: A survey,” ISPRS J. Photogramm. Remote Sens., vol.
196, no. April 2022, pp. 146–177, 2023, doi:
10.1016/j.isprsjprs.2022.12.021.
[10] E. Candela, O. Doustaly, L. Parada, F. Feng, Y. Demiris, and P.
Angeloudis, “Risk-aware controller for autonomous vehicles using
model-based collision prediction and reinforcement learning,”
Artif. Intell., vol. 320, p. 103923, 2023, doi:
10.1016/j.artint.2023.103923.
[11] F. V. Monteiro and P. Ioannou, “Safe autonomous lane changes
and impact on traffic flow in a connected vehicle environment,”
Transp. Res. Part C Emerg. Technol., vol. 151, no. March, p.
104138, 2023, doi: 10.1016/j.trc.2023.104138.
[12] Litman T (2015) Autonomous vehicle implementation predictions.
Victoria Transport Policy Institute 28.
[13] Bobbie D. Seppelt and John D. Lee "Keeping the driver in the loop:
Dynamic feedback to support appropriate use of imperfect vehicle control
automation",Volume 125, May 2019, Pages 66-80,
https://doi.org/10.1016/j.ijhcs.2018.12.009
[14] Padmaja, B., Moorthy, C.V.K.N.S.N., Venkateswarulu, N. et al.
Exploration of issues, challenges and latest developments in autonomous
cars. J Big Data 10, 61 (2023). https://doi.org/10.1186/s40537-023-00701-y
[15] Campbell M, Egerstedt M, How JP, Murray RM. Autonomous driving
in urban environments: approaches, lessons and challenges. Philosophical
Transactions of the Royal Society of London A: Mathematical, Physical and
Engineering Sciences. 2010;368(1928):4649–72.
[16] Abudayyeh, Dana, Malek Almomani, Omar Almomani, Hadeel Alsoud,
and Farah Alsalman. 2023. "Perceptions of Autonomous Vehicles:
A Case Study of Jordan" World Electric Vehicle Journal 14, no. 5:
133. https://doi.org/10.3390/wevj14050133
[17] Nick Gang, Srinath Sibi, Romain Michon, Brian K Mok, Chris Chafe,
and Wendy Ju. 2018. Don't Be Alarmed - Sonifying Autonomous
Vehicle Perception to Increase Situation Awareness..
In Proceedings of the 7th international conference on automotive
user interfaces and interactive vehicular applications. ACM Press,
New York, New York, USA, 237–246.
[18] Ioannis Politis, Stephen Brewster, and Frank Pollick. 2017. Using
Multimodal Displays to Signify Critical Handovers of Control to
Distracted Autonomous Car Drivers.IJMHCI 9, 3 (2017), 1–16.
[19] P. Kaur, S. Taghavi, Z. Tian and W. Shi, "A Survey on Simulators for
Testing Self-Driving Cars," 2021 Fourth International Conference
on Connected and Autonomous Driving (MetroCAD), Detroit, MI,
USA, 2021, pp. 62-70, doi:
10.1109/MetroCAD51599.2021.00018.
[20] Savari, Maryam, and Yoonsuck Choe. 2022. "Utilizing Human
Feedback in Autonomous Driving: Discrete vs.
Continuous" Machines 10, no. 8: 609.
https://doi.org/10.3390/machines10080609
