Page 205 - Kaleidoscope Academic Conference Proceedings 2021

Connecting physical and virtual worlds

handles all information regarding the communication aspect (position, orientation, acceleration, etc.) of each mobile
of the simulation, such as data traffic, buffer and channel entity. For this problem, the samples are collected at
information. In the CAVIAR simulations, channels can be every 10ms and they contain information of 37 entities (34
pre-computed and the communication simulation decoupled pedestrians, 2 cars, and 1 UAV).
from the physics engine, as often used in AI/ML applied to
beam selection [7, 4]. As shown in Figure 4 the spatial data generated by the
virtual scenario is the input for the CAVIAR simulation
The 3D assets used in the Environment and as Mobile entities, environment, more specifically, the communication engine,
such as UAVs, cars, buildings, etc, are either created or which is responsible for computing the radio channels and
obtained online, as described in [4]. They compose the other parameters related to the telecommunication system,
simulation environment as fixed or mobile objects, whose such as buffer size, etc. The output of the communication
eventual movements and interactions are managed by the engine along with the spatial data is the input for an RL
Mobility engine and by the Physics engine of the virtual agent, that is trained to choose, in each time slot, a user to
world subsystem, respectively. The Sensors engine output serve, and the beam that should be used.
constitutes the input to the AI/ML frontend engine.

The AI/ML engine receives signals and communication 3. COMMUNICATION MODEL
parameters that were simulated in the virtual-world
The simulation environment incorporates not only the
environment and suggests actions that are then implemented
RL-related functions required by the OpenAI Gym
by the Orchestrator, that also considers parameters from the
Application Programming Interface (API), but also the
Communications engine and Environment. An example of an
functions related to the communication system. This section
action in the context of beam selection would be sending to
will describe the adopted communication model.
the base station a list of codebook indices to try and avoid a
full beam sweeping.
We assume downlink transmission using a carrier frequency
f c = 60 GHz and a bandwidth of 100 MHz. The BS serves
three distinct receivers or UEs, which are located at a car, at
2.2 CAVIAR simulation for user scheduling and beam
a UAV and used by a pedestrian. The BS has an individual
selection problems
buffer with a size of 1 Gb for each receiver. We assume that
each packet has 8188 bytes. When the buffer of a specific
For the user scheduling and beam selection problems, the
user becomes full, the newly arrived packets are dropped.
Communications engine used by CAVIAR simulations was
defined by geometric channel models, further described in
The MIMO system corresponds to an analog architecture
Section 3. The Physics engine and the Mobility engine are
using a Uniform Planar Array (UPA) with N t antenna
handled by Unreal Engine and AirSim, and finally, for the
elements at the BS, and receivers with UPAs with N r
AI/ML engine, we assume an RL environment. Upon having
antennas. Therefore, the MIMO channel between the BS
the episodes available, the environment can be executed
and a given user is represented by a N r × N t matrix H. The
to allow an agent to assume the role of a BS, scheduling
codebooks are obtained from Discrete Fourier Transform
and serving the users with a specific beam chosen from its
} and
(DFT) matrices and denoted by C t = {¯ w 1 , · · · , ¯ w N t
codebook. The episodes can be used to train and test agents,
¯ ¯ }. They are used at the transmitter and
as well as store the outputs in a CSV file, which can be used C r = {f 1 , · · · , f N r
the receiver sides, respectively. When modeling the beam
as it is or reproduced graphically in Unreal, as described in
selection, the chosen beam pair [p, q] is represented by a
Figure 4.
unique index i ∈ {1, 2, · · · , M}, where M ≤ N t N r . Each
index p (or q) generates a specific radiation pattern, as
Unreal/AirSim
depicted in Figure 5 for a N t = 64.
Waypoint Episodes
generator For the i-th index, the equivalent channel is calculated as:
Simula on ∗ (1)
i
y i = w Hf i ,
environment
Rendering
ˆ
and the optimal beam index i is given by
Agent output file
(scheduled user and RL agent RL agent
i = arg
codebook index) test training ˆ max |y i |. (2)
i∈{1,···,M }
Figure 4 – CAVIAR simulation flow. We do not take noise into account in order to isolate the
impact of the beam selection procedure.
Using a virtual scenario provided by CAVIAR, three mobile
entities: a pedestrian, a car, and a UAV are simulated in Ray Tracing (RT) was used in [4], to generate realistic
order to generate a data set of urban mobility. This data communication channels H. For this paper, we have not used
is organized in episodes that contains spatial information RT but a simpler procedure based on the geometric MIMO
– 143 –