THE 3 rd INTERNATIONAL SCIENTIFIC CONFERENCES OF STUDENTS AND YOUNG RESEARCHERS dedicated to the 99
th
anniversary of the National Leader of Azerbaijan Heydar Aliyev
198
As navigation sensors may create real-time data streams at high data
rates (typically in the gigabits per second), cars can't send all of their sensor
data within the available bandwidth. One major finding is that different types
of data collected by sensors are of varying relevance. Carcel takes
advantage of this by instructing the cloud to seek information at a higher
resolution from only the most critical parts of the environment. These
requests are sent to the car via the cloud's request module. It analyzes
sensor data in aggregate, represented using the Octree data structure, and
creates requests for different locations at specific resolutions depending on
sensor data lacunae. The receiver module feeds sensor data to the planner
module, which collects data from different autonomous vehicle and static
road-side infrastructure sensors. The receiver module on each autonomous
vehicle registers requests and advises the transmitter to transmit a higher
proportion of sensor data from the requested locations at the appropriate
resolution.
[2] In overall, first we create a real-time connection pool between each
endpoint of the agents which generates multitude of data points to gather
information. We have applied multi-robot control systems over the
autonomous cars for simulating more practically and give solid output from
our hypothesis. That is why, our implementation on autonomous vehicles
consists of hardware sensor logging, data transmission and data receiving
from cloud platform. Additionally, the methodology contains model training
under the support of cloud infrastructure with GPU acceleration.
All things considered; the research paper suggests that cloud computing
is increasingly in high priority to rocket the efficiency of multi-robotics
systems and autonomous vehicles. In this paper, we have provided different
types of approaches and solutions with different methodologies by other
researchers as well as depicted some algorithmic scheme and diagrams to
illustrate more visually the architecture which is intended. Utilizing cloud
resources in equivalent load balancing techniques we can conclude with
better design perspective which is claimed mainly in our paper. Thus, many
agents that followed by cloud platform and were connected to the remote
server can transfer data and then benefit from the common data storage and
predictions in the cloud system.
References [1] S. Liu, J. Tang, C. Wang, Q. Wang, and Jean-Luc Gaudiot, (2017) “Implementing a Cloud
Platform for Autonomous Driving”, (pp 1-8) Fellow, IEEE
[2] J. Levinson, J. Askeland, J. Becker, J. Dolson, D. Held, S. Kammel, J. Kolter, D. Langer,
O. Pink, V. Pratt, M. Sokolsky, G. Stanek, D. Stavens, A. Teichman, M. Werling, and S.
Thrun. (2011) “Towards fully autonomous driving: Systems and algorithms”. (pp 163-168)