Ready, trainer … one*! discovering the entanglement of adaptive learning with virtual reality in industrial training: a case study

Figures & data

Figure 1. Images of the current physical door simulator.

Figure 2. Adaptive learning boundaries embed in the “flow channel” (adapted from Göbel and Wendel (2016) and Paramythis and Loidl-reisinger (2004)).

Figure 3. Followed design process adapted from Peffers et al. (2007) with numbers indicating respective sections and subsections of the paper.

Table 1. Group characteristics.

Participant	1	2	3	4	5	6	7	8	9	10
Age  (years)	29	64	43	44	63	63	51	50	48	49
Experience in the Railway Industry  (years)	2	30	22	20	18	19	8	4	5	2
VR Experience  (instances of previous use)	1	1	1	0	0	1	0	0	0	0

Table 2. Solution objectives and requirements.

Objective	Requirements
The virtual simulator achieves at least the same scores on learning preference and learning outcome as the current door simulator	The simulator features high levels of physical, semantic, and phenomenological realism The simulator has added value to the conductors The simulator limits anxiety The simulator is focused on obtaining the correct declarative and procedural knowledge
The virtual simulator provides an adaptive learning experience	The simulator has multiple entrance points The simulator measures skill level and progress The simulator adapts its recommended learning steps towards the skill level of the student The application is a static intrinsic game The application has a storyline
The virtual simulator provides logistical benefits over the current simulator	All costs associated with setting up the virtual door simulator in one location are less than 80.000 The virtual simulator requires maintenance or support less than four times a year Conductors can use the simulator without an instructor present
The virtual simulator allows for experimentation and collaboration to acquire new knowledge	Conductors can work together if that is required in some departure scenarios. The VR solution allows participants to observe the performance of others The solution limits the cognitive load to a minimum Interaction with VR should be natural and not in the way of learning. The virtual simulator includes multiple train types to practice on
The virtual simulator provides an adequate amount of technological immersion	Users are adequately represented in the VLE The simulator features spatial audio The visuals are physically and semantically realistic The objects in the VLE behave consistently Interactions in the VLE are embodied
The virtual simulator provides a high amount of task immersion and an adequate amount of social immersion	The tasks should be simulated as authentic as possible The simulation features deep levels of manipulation The VR-simulator incorporates the bigger picture of the departure processes: Travellers Platforms Time pressure Safety hazards

Figure 4. Holistic overview of the design and development process.

Table 3. Results of the conductor workshop on the hardware concepts.

Concept	Strong points	Weak points	Preference
Standalone	Use both hands Most possibilities in terms of interactions	Complicated interactions Hygiene	2
Mobile	Simple controller	Just one controller Unpractical during conductor training, switching phones, etc. Hygiene	1
Kinect	Natural interaction without controllers Fun experience	Always facing screen No real environment The sensor does not easily pick-up a new player after switching	2

Table 4. Results of the conductor workshop on the software concepts.

Concept	Functionality improvements	Comments	Preference
Basic	Record practice so that replays can be watched Including the smartwatch	Looks like a well-founded basis Think about the used terminology. It should match the slangs conductors use on the job	5
Multiplayer	Include communication via a transceiver	Seems essential because it is also possible in the current simulator	5
Adaptive learning	Do not use a star system	Conductors do not like to be assessed during practice	2
Serious gaming	Also, practice other parts of the job	I would like to use it at home to show my friends and family	2

Table 5. Results of the multi-criteria analysis.

Criteria	Sub criteria	Definition	Weighting factor	Concept standalone	Concept mobile	Concept Kinect	Current simulator
Costs	Development costs, Hardware costs, maintenance costs	Costs of hardware only, not development time	4	8	9	7	1
Learning curve/complexity	null	How difficult is it learning to use the hardware in such a way that you can practice properly	3	3	6	5	10
Use setting flexibility	Learning content, Location,	In how many contexts or locations can the hardware be used	5	7	9	9	2
Digital quality of the simulation	Visual, Audio, Feeling	How well does the simulation physically represent reality	4	7	4	3	4
True to nature virtual object manipulation	Null	Is interacting with virtual objects the same as you would do with those objects in the real world?	5	7	2	9	10
True to nature navigation	Null	Do you navigate through the virtual world the same way as you would in the real world	4	10	7	4	10
Comfort of experience	Motion sickness, physical comfort, eye tiring	How long can you pleasantly use the hardware for practice	3	3	7	8	10
Maintenance & support	Repair, Replace, Preventive maintenance	How well is the hardware supported by the manufacturer and how difficult is it to maintain in case of malfunctions	4	7	7	1	7
TOTAL				216	202	189	208

Figure 5. Activity block diagram of the application.

Figure 6. Functional block diagram of the application.

Figure 7. Paper prototype.

Table 6. Detailed functionality of the VR training system.

Function	Subfunctions
Loading scenario and user avatar	Representing the player’s body Showing a complete train and train station environment and travellers Allowing walking around of the player
Allowing manipulation of all required parts	Allowing turning of the key Allowing pressing of door buttons Allowing converting of switches Allowing closing door by hand Allowing removing the cover of the door mechanism
Keeping track of mistakes	Listing tasks to be done during each scenario Listing tasks that are done during the execution of the scenario
Providing feedback	Comparing the two task lists Listing the missing tasks Listing tasks that were mixed up Explaining why missing tasks are important and how they need to be done
Showing main menu, train choice menu, or scenario choice menu	Showing clickable buttons Showing controllers and pointers to click Linking buttons to correct menu

Figure 8. Design logic of the adaptive base level test.

Figure 9. Screenshots of the final version of the prototype.

Figure 10. Schedule for evaluation sessions.

Table 7. The average time needed, and the average number of tips asked per exercise by the 10 participants, with one standard deviation.

Exercise	Average Time (mm: ss)	Average Number of Tips
1	03:08 ± 00:50	1.3 ± 0.8
2	06:52 ± 1:35	2.1 ± 0.7
3	13:01 ± 2:50	3.8 ± 1.2

Figure 11. Evaluation of the educational impact of both solutions.

Figure 12. Overall learning preference of the conductors with both solutions.

Table

Price:	€449 (64GB) €549 (128GB)
Field of View (Horizontal)	90°
Resolution	1440 x 1600 per eye
Weight	570g
Battery time	2-3 hours
Charging time	2 hours
Speakers	Built-in
# controllers	2

Table

Price	€120
Field of View (horizontal)	100°
Resolution	Dependent on smartphone
Weight	260 grams + weight of smartphone
Smartphone size	4,7” - 6”
Smartphone requirements	60 Hz refresh rate, Full HD display, Dual core processor
Speakers	Smartphone
# Controllers	1

Table

Price	€1500-2500 (Kinect + PC)
Resolution	Preferably 4K
Display size	To be decided
Sensor range	50cm – 450cm
FOV horizontal	70°
FOV vertical	60°
Max number of people	6
Speech recognition	Yes

Table

Expected benefits of VR	Number of responses (respondents could choose multiple benefits) N=10
More realistic scenarios	9
Not having to go to a training point	5
More or flexible learning content	5
Practice without instructor	4
Practice quickly when you want	1

Table

Scenario	Comments	Observations
Regular departure	It does feel a little bit like a game The environment is highly realistic Some people prefer sitting down because of fear of instability/nausea	Turning the key can be troublesome They are primarily struggling with the controls/remembering which buttons to use
Lock from outside	They would like to practice communication using a transceiver too They have to walk too far to the train cabin. In reality, they would choose a different door to do this procedure on	Using controllers is going significantly smoother. After having practiced the first scenario, this one poses little trouble Many still ask what steps to do, even though they know the procedure by heart
Door malfunction	Time pressure could be something to make this scenario more realistic The scenario represents the difficulties conductors face in this situation quite well	Goes well except for using the key at heights There seem to be different opinions on how to solve a door malfunction: the goal of this scenario is not clear for some

Göbel, S., & Wendel, V. (2016). Personalization and adaptation. In R. Dörner, S. Göbel, W. Effelsberg, & J. Wiemeyer (Eds.), Serious games: Foundations, concepts and practice (pp. 161–210). Springer International Publishing.

 Google Scholar

Paramythis, A., & Loidl-reisinger, S. (2004). Adaptive learning environments and e-learning standards. Electronic Journal on e-Learning, 2(1), 181–194.

 Google Scholar

Peffers, K., Tuunanen, T., Rothenberger, M. A., & Chatterjee, S. (2007). A design science research methodology for information systems research. Journal of Management Information Systems, 24(3), 45–77. https://doi.org/10.2753/MIS0742-1222240302

 Web of Science ®Google Scholar