IEEE Robotics & Automation Magazine - March 2021 - 75

robustness. These make the system effectively cope with
highly dynamic indoor environments, such as hospital
wards, with few estimation errors and a low computational burden [18]. More sophisticated SLAM approaches
can be found in the literature [19]; however, they suffer
from drawbacks that make them unsuitable for this
application. For example, visual SLAM methods may
suffer from the invalidity of the static world assumption, which limits the application of RGB-D SLAM in
dynamic environments [20].
Preliminary Autonomous Navigation Evaluation
The autonomous navigation system was preliminarily validated in laboratory settings that reproduced the healthcare ward. This phase facilitated measuring the system's
performance and assessing the reliability and robustness of
the navigation system, especially in the presence of static

Table 1. The navigation parameters implemented
on the UCBM COVID robotic platform.
Global Planner Parameters

Value

Update frequency of the global planned path

1 Hz

Neutral cost

Cost factor

Lethal cost

253

Tolerance at the goal point

0.1 m

Local Planner Parameters

Value

Time to forward-simulate trajectories

1.7 s

Update frequency of the local planned path

1 Hz

Controller frequency

10 Hz

Explored samples in the x velocity space

Explored samples in the y velocity space

Explored samples in the theta velocity space

Step size to take between points on a given trajectory

0.025 m

Weight for how much the controller should stay
close to the path

Weight for how much the controller should
attempt to reach its local goal

Weight for how much the controller should
attempt to avoid obstacles

0.01

Tolerance in reaching goal position

0.2 m

Tolerance in reaching goal orientation

0.2 rad

Maximum transnational velocity

1.5 m/s

Maximum rotational velocity

2 rad/s

Cost Map Parameters

Value

Global and local cost map update frequency

10 Hz

Local cost map obstacle range

3.5 m

Local cost map ray trace range

Local cost map width and height

Local cost map resolution

0.025 m

and dynamic obstacles. Two controlled environments were
mapped, and the obstacle-avoidance robustness was tested.
The first environment, shown in Figure 3(a), represents an
open space where the robot can navigate autonomously.
The latter, reported in Figure 3(b), is a more complex environment, consisting of a narrow passage between two
areas. The robot could reach one area by crossing a
1.2-m-wide corridor resembling the COVID treatment
center where the robot was intended to operate. The tests
were conducted under three conditions: the first aimed at
testing the repeatability of the performed trajectory in an
undisturbed condition, the second assessed how the
implemented autonomous navigation handled the presence of static obstacles, and the third included the presence of dynamic obstacles (e.g., a health-care provider
crossing in front of the robot).
The robot was directed to autonomously navigate from
the starting pose X 0 toward the target point X 1 and vice
versa. The reference fixed frames are reported in Figure 3(a) and (b) for the two environments used for the
platform testing. Position and orientation are expressed
by means of Cartesian coordinates and quaternions,
respectively. Poses X 0 and X 1 for each environment are
presented in Table 2. The robot performed each navigation task 16 times in each proposed environment. To validate the algorithm in the presence of static obstacles, a
box with dimensions of 0.3 # 0.21 # 0.5 m was placed in
p obstacle, expressed in the reference frame O - xyz, specifiE1
cally, p obstacle
= [1.5, 0, 0] T m in environment E1 and
E2
p obstacle = [1.45, 2.25, 0] T m in E2. The robot was asked to

y p0
0z x

y 0
zx p
0

(a)

(b)

Figure 3. Maps built by the robot in the laboratory setting. The
fixed reference frame O - xyz and the target points p0 and p1
used to validate the robot's navigation capability are highlighted
for the two tested environments. (a) The open-space scenario.
(b) The more challenging indoor scenario.

Table 2. The target poses used in the navigation
capability test phase.
E1

Pose

Position (m)

Orientation ( )

p 0 = [0, 0, 0] T

q 0 = [0, 0, 0, 1] T

p 1 = [2.5, 0, 0] T

q 1 = [0, 0, 0, 1] T

p 0 = [0.53, - 0.42, 0] T

q 0 = [0, 0, - 0.70, 0.71] T

p 1 = [3.92, - 0.42, 0] T

q 1 = [0, 0, - 0.77, 0.64] T

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