IEEE Robotics & Automation Magazine - March 2021 - 63

measures analysis of variance (ANOVA) design study with two
factors: irradiation condition (robot versus static for a total time
of 30 min) and time (five levels with color sampling performed
every 6 min).
Results
Figure 3 reports the distribution of irradiation at the end of the
experiment in the room. After 30 min of exposure, log-3
SARS-CoV-2 virus inactivation values (3.7 mJ/cm2 [8])
were reached by 100% of the samples in the dynamic
condition and by only 52.38% of the samples (11/21) in the
static condition (Fisher exact test, p = 0.0005). The F-statistics associated with the two-way ANOVA experimental
design revealed a significant effect for both factor time
(F (1, 4) = 33.51, (p 1 1e -5) and the interaction factor " time
× experimental condition " (F (1, 4) = 2.876, p = 0.028).
Moreover, a difference in the treatment condition was obtained, although still nonstatistically significant, at 0.05 level
(F (1, 4) = 3.455, p = 0.078). The results clearly indicate
that there is an interaction between time and experimental
condition; the exposure to radiation was strictly dependent
on time in the static condition (linear relation of released
energy), and, when the robot worked near the static source
location, its effect on the markers' irradiation was close to
the effect of the static source.
Differences in the absorbed energy of nearby markers,
especially in the static condition, evidence the effect of
object shadowing (i.e., chair armrests and backseats) and of
the different surface orientations. The contribution of indirect radiation by light diffusion appears negligible in the
explored exposition time. Figure 3 (c) displays the mean
value of the delivered UV-C irradiation energy over time
(the " Average Absorbed Energy " graph), computed as the
mean of all detectors at each time, with higher values of
irradiation delivered in the dynamic condition compared to
the static one. As illustrated in Table 1, at 24 min a significant difference in delivered irradiation energy (paired t-test
p 1 0.05) was reached, with almost the same level kept at
30 min. In fact, if the robot works mostly in proximity of
the static lamp's position during a time interval, the mean
difference remains almost the same, as the two conditions
overlap considerably.
In this experiment, the two limit conditions of a static
source of radiation compared to a continuously moving
one were tested. However, in a real application of a static
UV-C lamp, a human operator could manually reposition a
fixed radiation source several times to increase the uniformity of the energy absorption. To better understand the
effect of this operation, we have simulated the distribution
of the absorbed energy along a vertical wall at an increasing
number of repositionings of a radiating source held at a
fixed distance from the wall. The results in Figure 5 present
how the level of homogeneity of energy absorption depends
on the number of repositionings and on the distance of the
lamp from the wall. The peaks can be clearly observed in correspondence to the fixed positions of the lamp [Figure 5(a)]

while an almost flat profile with some boundary effects is
achieved in the case of a moving source, obtaining an optimal distribution of the irradiated energy along all of the
surface. Correspondingly, in Figure 5(b) the simulation
demonstrates how the
required exposure time
After 30 min of exposure,
increases if the same
minimum level of irradilog-3 SARS-CoV-2 virus
ation energy is requested.
In particular, when the
inactivation values
geometry of a room, i.e., a
rectangular or any asym(3.7 mJ/cm2) were reached
metrically shaped room
(such as a corridor), or
by 100% of the samples in
obstacles within a room
constrain the robot to be
the dynamic condition.
close to the surface, the
time required by a fixed
lamp must be a multiple
of the time required by a moving one. If we consider the cost
of the manpower needed to reposition the source, overall disinfection costs are then multiplied by a higher value.
A Motion-Planning Algorithm Based on
Irradiation Physics
Overview
This first study demonstrated how robots can improve disinfection accuracy and reduce task-execution time by moving
the radiation source, thus distributing energy more homogeneously in the environment. The disinfection problem can be
seen as a special case of motion planning and navigation in
dynamic environments, where the mission target is composed
of a large number of variable target points, i.e., all of the

Color
Green Value (RGB)
Energy (mJ/cm2)

248 181.2
0

174.7

25

50

141.1 117.7
75

100

Figure 4. The reference chart provided by the manufacturer
for the energy-density calculation from the G channel of the
marker's color.

Table 1. The paired t-test of a UV-irradiation
average dose, * p < 0.05.
Time

Mean

Standard
Deviation

p Value

6 min

3.22

15.09

Not significant

12 min

7.22

23.66

Not significant

18 min

10.84

27.98

p = 0.091

24 min

13.21*

28.42*

p < 0.05

30 min

13.21

30.45

p = 0.061

MARCH 2021

*

IEEE ROBOTICS & AUTOMATION MAGAZINE

*

63
IEEE Robotics & Automation Magazine - March 2021

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