IEEE Robotics & Automation Magazine - March 2023 - 32

participants in UCL's PAMELA (Pedestrian Accessibility
Movement Environment Laboratory) facility. PAMELA is
equipped with a modular platform that can be used to replicate
gentle slopes (around 10º) and negative obstacles (with a
roughly 30-cm drop). We constructed the circuit shown in
Figure 11, which consists of static components (e.g., a door, a
narrow passageway, and an elevator) that have been identified
as relevant for this case study by clinicians [6]. The circuit is
composed of lightweight cardboard sheets to ensure participants'
safety in case of collisions. The absolute position of the
moving wheelchair is estimated with a vision system based
on multiple cameras attached to the ceiling. The encouraging
results of these trials, have been recently presented in [27].
TRAINING IN VR
The wheelchair simulator described in the " Wheelchair Simulator "
section provides a high-fidelity, 3D immersive environment,
and it offers the possibility to repeat the same
circuit multiple times under identical experimental conditions.
The user gets the impression of driving a real wheelchair,
and safe navigation is guaranteed at all times. Shorter
training sessions are thus necessary, and a wider array of
(indoor/outdoor, obstacle-free/cluttered) environments and
real-life conditions (variable light conditions, moving pedestrians)
can be tested.
Driving a real wheelchair could be dangerous for people
with disabilities, requiring extensive training sessions to
acquire the ability to move safely. The goal of the clinical
trial SIMADAPT1 (NCT04171973) was to verify whether the
performance observed on a real circuit was comparable to
the one experienced on the wheelchair simulator. To this end,
the wheelchair users were asked to complete the three obstacle
courses considered in SWADAPT1 and SWADAPT2 (see
the " Driving Assistance " section) in the real world and in the
VR environment. In total, 29 expert drivers with neurological
degenerative disorders were screened by clinicians to take
part in this study. The results show that there is no statistically
significant difference between the real world and VR (Kruskal-Wallis
test). Participants' quality of experience (QoE) was
measured via a USE (Usefulness, Satisfaction, and Ease of
Use) questionnaire with 30 questions grouped into four criteria
and rated on a seven-point Likert scale [28], as reported in
Figure 12. In addition, if the cognitive load is generally higher
in VR, the VR/real-world cognitive load ratio decreases as the
difficulty of the circuits tested by the users increases. In VR,
the patients experienced a high sense of presence, and the level
of cybersickness remained very low in the three circuits. In
particular, the collected data indicate that using the simulator
during a training phase could drastically reduce damage
to the environment (walls, doors, and furniture) and driving
accidents [18].
The objective of the clinical trial SIMADAPT2
No
13%
Yes
87%
(a)
27
20
13
7
Yes
83%
(b)
No
17%
(NCT04894981) was to evaluate the impact of the immersive
environment on VR driving performance. Three different
conditions were compared in SIMADAPT2: with a Cave
Automatic Virtual Environment (Immersia at IRISA/Inria
Rennes), with a head-mounted display,
and with a nonimmersive TV screen
(see Figure 4). Overall, 18 wheelchair
users with and without driving difficulties
participated in this clinical study,
organized in two sessions to comply
with COVID-19 restrictions. Similar to
SIMADAPT1, our preliminary results
consistently show a small simulatedto-real
gap, strong acceptability and a
feeling of safety, and better driving performance
with the immersive displays.
Again, our data support the idea that
training with the simulator during a
learning phase leads to a significant
reduction of damage to property.
No
37%
Yes
63%
(c)
DISCUSSION: CHALLENGES AND
RECOMMENDATIONS
% of Time
(c)
FIGURE 10. The statistical results from the questionnaire used to evaluate SpheriCol [21].
The (a) ease of learning, (b) ease of use, and (c) usefulness of SpheriCol. The (d)
percentage of time that driving assistance was used by the 17 participants in the test
circuit. NA: not available.
32 IEEE ROBOTICS & AUTOMATION MAGAZINE MARCH 2023
TECHNICAL CHALLENGES
An open challenge is to guarantee that
the ensemble of assistive technologies
developed in the ADAPT project by
the French and English partners works
safely and harmoniously.
A possible way forward is to exploit
redundant information. For instance,
% of Participants
100
90
80
70
50-30
20-10
NA

IEEE Robotics & Automation Magazine - March 2023

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - March 2023