IEEE Robotics & Automation Magazine - March 2021 - 79

The deviation with respect to the linear path was very
small and can be seen in Figure 7(c).
The second scenario, illustrated in Figure 3(b), was also
used to test the robot's navigation capability. The robot's
paths are reported in Figure 7(d)-(f). In the unperturbed
condition, the robot was able to reach the target position
in 21.32 ± 1.02 s along a path 7.79 ± 0.19 m long. The target was reached in all the trials with a DE of 0.04 ± 0.02 m.
The introduction of the static obstacle led the robot to take
a longer path to reach the assigned position. The CT consequently increased. The presence of the static obstacle did
not affect the capability of the robot to reach the target,
with a small residual error. In fact, the resulting DE was
0.05 ± 0.02 m. During the third condition, i.e., a person
crossing the robot's path, the robot's replanning capabilities
were assessed. As expected, the robot stopped when a
dynamic obstacle appeared in front of the mobile base.
Therefore, the mean CT was much greater than those measured in the other experimental conditions (39.06 ± 14.64 s).
The PL computed in this case was 8.68 ± 2.24 m. The high
variability shown in these trials results from the fact that
the robot, in reaching the target in position p 1, searched
for different paths when the dynamic obstacle obstructed
the corridor, thus lengthening the PL. On the other hand,
the robot was capable of reaching the assigned target with
a small DE (0.06 ± 0.03 m).
All the tasks performed by the robot obtained a high
SR. The robot was always able to avoid collisions with the
obstacles and to reach the assigned target position with a
small DE. The SR was less than 100% due to some DEs
greater than 0.1 m that occurred in a few trials, i.e., 8%,
corresponding to eight of the 96 trials. The results demonstrate that the proposed system was able to complete the
required tasks by safely managing (both for the robot and
for the user) unexpected events, such as the presence of
static and dynamic obstacles.
After this validation in a laboratory setting, the robot
was installed at the UCBM COVID center at the beginning
of May 2020. The main results collected during two
months of use are reported and discussed in this section.
Figure 8(a) conveys the map generated by the robot during
the mapping phase. The red dots represent the PoIs listed
in the previous section, while the blue dashed lines are the
trajectories the robot executed during a daily logistics
function. Quantitative results for the logistics scenario are
summarized in Table 4. Mean values, computed during the
total number of days of use, are reported with their standard deviation.
The robot worked in the logistics scenario for 43 days,
for roughly 7 h per day, from 8 a.m. to approximately
8 p.m. During the daily activities, at least two recharge phases were required. The average traveled distance inside the
ward, where the robot transported drugs, tubes, blood
components, and medical devices, was roughly 8,000 m
per day. The number of human operators required for the
logistics scenario was three. The operators were inside the

hot cell, the red area, and the intensive short observation
and holding areas. Each user filled and emptied the robot's
cargo box according to their needs and determined the
robot's target location. In this way, the operators did not
have to leave their work station while the robot transported materials.
Figure 8(b) provides a UCBM COVID treatment center
map with the trajectories executed by the robot during a
disinfection procedure. As explained in the previous

Hot
Cell

OBI
Area

Red
Area

Charging
Station

Holding
Area

(a)

(b)
Figure 8. A map of the UCBM COVID treatment center that the
robot built. In (a), red points represent the targets the robot had
to reach for the logistics scenario. The dashed blue lines indicate
trajectories the robot performed during a day. In (b), the dots
represent targets the robot had to reach for the disinfection
scenario. The two dotted lines, in orange and blue, represent
the robot's paths to treat the entire hallway. In particular, the
points where the robot stopped to treat the environment are
highlighted with large dots.

Table 4. The results of the logistics scenario.
Scenario 1: Logistics
Days of use

43 days

Mean time delivering materials

437.7 ± 28.2 min

Number of recharges per day

2.13 ± 0.74

Mean traveled distance

7,971.2 ± 44.1 m

Number of health-care operators

Usage indexes of the proposed robotic platform are reported as the
mean value and the standard deviation.

MARCH 2021

IEEE ROBOTICS & AUTOMATION MAGAZINE

IEEE Robotics & Automation Magazine - March 2021

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