IEEE Robotics & Automation Magazine - March 2021 - 68

Methods
The room was instrumented with 24 colorimetric UV-Csensitive markers with different positions and orientations,
located on tables (1-4), on shelves (5-9, 11, 14, 22), on
walls (10, 12), on the door (13), and on the floor (15-21,
23, 24) (see Figure 9), and then it was exposed to UV-C
irradiation. To have a benchmark for evaluating the algorithm performance, it was necessary to define a comparative trajector y that
ensured good coverage
Shared environments, such of the room area and,
thus, offered a fair comas conference rooms, public parison. We then chose
the simplest trajectory
spaces, or workplaces,
allowing the robot to
move across the room at
could benefit from robotic
a constant distance from
the objects to disinfect
disinfection in terms of
(we call it the " standard "
trajector y), and the
task accuracy, cost, and
result is the path displayed in Figure 9(a),
execution time.
traveled at a constant
speed. Then, we performed the irradiation
experiment in two different conditions: 1) the " standard "
trajectory and b) a GA-optimized trajectory. The experimental comparison was carried out with equal total irradiation time for the two conditions (resulting also in equal
total energy irradiated in the room). For this reason, we
calculated the speed for the " standard " trajectory to match
the total time needed for the GA-optimized trajectory: 594
seconds. The energy data were evaluated with the same
method described in the previous " Methods " section.
Results
Figure 9 displays the results of the experiments in the two
different conditions. The bar plot of the markers' energy-
density distributions shows that, in the " standard " condition, only 18/24 markers were successfully disinfected
(energy density > 16.9 mJ/cm2) while, in the GA condition,
the score was 24/24 (Fisher exact test p = 0.022). Although
the average delivered energy in the GA condition is higher
than in the " standard " condition and the standard deviation is lower in the GA condition than in the " standard "
condition, the number of samples was not high enough to
achieve a statically significant difference between the two
conditions. Numerical results are reported in Table 3.
Discussion and Conclusions
This article presents 1) the performance evaluation of a
UV-C-disinfection mobile robot, compared to conventional disinfection based on static UV-C lamps, and 2) the
evaluation of a new trajectory-planning strategy specifically for disinfection robots. The evaluation was performed using UV-C-sensitive colorimetric detectors to
68

IEEE ROBOTICS & AUTOMATION MAGAZINE

MARCH 2021

measure the effective amount of absorbed energy; then
data were elaborated to assess statistical significance. The
static versus dynamic experiment results demonstrate
that, in a wide environment, it is possible to improve the
effectiveness and reduce the time of the disinfection by
using a mobile source of UV-C irradiation. A significant
statistical difference on the number of disinfected markers
was achieved, which is the most important goal of the disinfection task.
However, this on-field experience highlights the importance of rearranging an environment to make it suitable
for robotic disinfection. The robot could not reach and
disinfect some areas of the room because they were too
distant or not sufficiently exposed. To improve the effectiveness of robot disinfection, environments should be
adequately prepared. There should not be narrow passages or obstacles left on the floor, and the most likely
contaminated objects or surfaces, such as handles,
tools, and desks, should be directly exposed to radiations.
Disinfection is a repetitive, frequent, and time-consuming task, and, given the demonstrated importance of a
mobile source of radiation, the development of mobile
robotic platforms for disinfection is a straightforward
solution. It can, alone, have a significant impact on the
effectiveness, time, and resources needed for the disinfection procedure.
Still, the potential of a mobile robotic platform can go
further than the movement of the source of UV-C radiation.
It allows the development of optimized trajectories that,
although more complex, can be followed with the same
accuracy and repeatability. We explored this opportunity
by developing a novel trajectory planner specific for the
disinfection task based on an APF, an iterative simulation
method based on irradiation physics, and optimization
through GA to find the most suitable trajectory. The trajectory planner has been tested in simulation, where it
could produce a suitable solution ensuring the completion
of disinfection for different test environments.
Still, improvements are needed to address cluttered and
complex environments, where the method could suffer
from the presence of multiple obstacles or nonconvex
boundaries. These environments can be addressed by
dividing them into suitable subenvironments that can
be handled by the proposed trajectory planner and by
using high-level logic to switch from one subenvironment to another.
Finally, by means of our UV-C mobile robot, we
experimented on the effectiveness of the trajectory planner in a real setting. Results, again, confirmed the
importance of carefully choosing the trajectory to minimize the time needed for disinfection. The proposed
model scored a 100% disinfection success rate (24/24
markers) against the 75% of the benchmark " standard "
trajectory; the benchmark trajectory requires more
time for completing the disinfection on all of the surfaces in the environment.

IEEE Robotics & Automation Magazine - March 2021

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - March 2021