TY - GEN
T1 - IoT Edge Server ROS Node Allocation Method for Multi-SLAM on Many-core
AU - Fukui, Masato
AU - Ishiwata, Yoichi
AU - Ohkawa, Takeshi
AU - Sugaya, Midori
N1 - Publisher Copyright:
© 2022 IEEE.
PY - 2022
Y1 - 2022
N2 - One of the most expecting IoT services is using multiple autonomous mobile robots. Drone and robots on nurse care whose services are expected to become widespread. For these robots, it is essential to create a map of the surroundings using Simultaneous Localization and Mapping (SLAM) technology. SLAM has a high processing load. Therefore, the high-performance edge servers will be placed close to the robot to support the computation. SLAM implementations typically consist of multiple software processes such as distance estimation and mapping. From this, it is considered that using a many-core processor for the edge server that can efficiently support SLAM for multiple robots. However, the effectiveness of using many-core processors for the IoT edge servers is not clear. Therefore, this study investigates the use of a many-core processor when SLAM is the target IoT application. To build a general-purpose mechanism for utilizing a many-core processor, we need a measurement system inside the ROS (Robot Operating System) which is the base middleware for SLAM, in order to measure the processing activity where the ROS components are asynchronously executed. Therefore, we developed a mechanism that makes it possible to accurately measure the execution of the components on ROS, and overcome the problem to allocate the many-core CPU for each ROS component execution properly. As a result, furthermore by considering the cache and properly allocating the ROS node to the processor, the execution performance was improved by up to 7.5%.
AB - One of the most expecting IoT services is using multiple autonomous mobile robots. Drone and robots on nurse care whose services are expected to become widespread. For these robots, it is essential to create a map of the surroundings using Simultaneous Localization and Mapping (SLAM) technology. SLAM has a high processing load. Therefore, the high-performance edge servers will be placed close to the robot to support the computation. SLAM implementations typically consist of multiple software processes such as distance estimation and mapping. From this, it is considered that using a many-core processor for the edge server that can efficiently support SLAM for multiple robots. However, the effectiveness of using many-core processors for the IoT edge servers is not clear. Therefore, this study investigates the use of a many-core processor when SLAM is the target IoT application. To build a general-purpose mechanism for utilizing a many-core processor, we need a measurement system inside the ROS (Robot Operating System) which is the base middleware for SLAM, in order to measure the processing activity where the ROS components are asynchronously executed. Therefore, we developed a mechanism that makes it possible to accurately measure the execution of the components on ROS, and overcome the problem to allocate the many-core CPU for each ROS component execution properly. As a result, furthermore by considering the cache and properly allocating the ROS node to the processor, the execution performance was improved by up to 7.5%.
KW - Edge computing
KW - many-core
KW - ROS
KW - SLAM
UR - http://www.scopus.com/inward/record.url?scp=85130628220&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85130628220&partnerID=8YFLogxK
U2 - 10.1109/PerComWorkshops53856.2022.9767431
DO - 10.1109/PerComWorkshops53856.2022.9767431
M3 - Conference contribution
AN - SCOPUS:85130628220
T3 - 2022 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events, PerCom Workshops 2022
SP - 421
EP - 426
BT - 2022 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events, PerCom Workshops 2022
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2022 IEEE International Conference on Pervasive Computing and Communications Workshops and other Affiliated Events, PerCom Workshops 2022
Y2 - 21 March 2022 through 25 March 2022
ER -