IEEE Robotics & Automation Magazine - March 2023 - 33

today, a growing number of accurate
3D models of indoor and outdoor environments
is publicly available. These
(CAD or point cloud) models could be
used in conjunction with spherical photogrammetry
(especially in areas that
the wheelchair user has never visited
before) to underpin real-time visionbased
motion estimation algorithms.
A first step in this direction has been
taken in [29], where a new panoramic
3D pose-tracking algorithm has been
shown to provide accurate estimates,
even in the presence of large interframe
motions (several meters). The
algorithm relies on a representation
of catadioptric images as a mixture of
photometric potentials, similar to the
one used for the direct visual gyroscope
in [22]. In future work, we plan
to adapt the approach in [29] to dualfisheye
images to have the largest
possible number of algorithms working
with the same hardware onboard
the wheelchair.
To guarantee safety, an assistive
Corridor Obstacle
Slope
Elevator
technology is also expected to operate
as designed in any circumstances
(including unfavorable conditions, such
as variable lighting, rain, uneven terrain,
and so on). Vision-based driving
assistance systems, such as SpheriCol,
might perform poorly in scenes with a
large range of light intensities, i.e., in
scenes where bright sunshine coexists
with dark shadows, as in the transitions
between indoor and outdoor environments.
Unfortunately, the price to pay for compactness in
consumer-grade twin-fisheye cameras is the limited dynamic
range. Real-time high-dynamic-range (HDR) vision will
finally make it possible to design assistive devices that work
in scenes with challenging illumination conditions. UPJV's
group has recently developed a new panoramic system (consisting
of an orthographic camera combined with four convex
mirrors and three neutral density filters), HDROmni [30],
which optically extends the dynamic range of the images. Preliminary
tests on a mobile robot are promising, and there are
plans to apply the same optical design to SpheriCol to make it
more robust to abrupt illumination changes. Another direction
for future research pertains to vision-based closed-loop control
and, in particular, heading control, for which a twin-fisheye
camera can be regarded as a valid alternative to conventional
MEMS gyroscopes integrated into the smart wheelchair.
The SDP approach to shared control holds great potential
(a)
for matching assistance to different driving styles. However,
while the computational heavy lifting is carried out offline,
Negative
Obstacle
(b)
FIGURE 11. UCL's PAMELA facility. (a) The modular platform enables setting up slopes and
negative obstacles. (c) The circuit, with a volunteer testing the model-based shared control
algorithm proposed in [26].
±0.98
±0.45
7
6
5
4
3
2
1
Ease of
Learning
±0.54
±0.95
±1.08
±1.4
±1.66
±1.44
Ease
of Use
Satisfaction
Usefulness
FIGURE 12. Evaluation of the QoE of the wheelchair simulator.
Mean and standard deviation of the USE score in the real world
(red) and the VR environment (blue), according to four criteria [18].
MARCH 2023 IEEE ROBOTICS & AUTOMATION MAGAZINE
33
Score
6.76
6.37
6.32
5.79
5.85
5.32
5.67
4.71

IEEE Robotics & Automation Magazine - March 2023

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