Public Abstract

DE-SC0019614: EdgeSense : Smart Sensors and FPGA based Heterogeneous Edge Computing Unit

Award Status: Inactive

Institution: CIRRUS360, RICHARDSON, TX
UEI: N/A
DUNS: 081510439

Most Recent Award Date: 02/12/2019
Number of Support Periods: 1
PM: Ndousse-Fetter, Thomas

Current Budget Period: 02/19/2019 - 09/18/2019
Current Project Period: 02/19/2019 - 09/18/2019
PI: SEN, SUDIPTA

Supplement Budget Period: N/A

Public Abstract

EdgeSense : Smart Sensors and FPGA based Heterogeneous Edge Computing Unit—CIRRUS360, 3017 Brookvale Drive, Richardson, TX 75082 Sudipta Sen, Principal Investigator, sudipta@cirrus3sixty.com Sudipta Sen, Business Official, sudipta@cirrus3sixty.com Amount: $213,255 Data influx from sources at the edge of the network is exploding e.g. from sensors to collect weather data, satellite imagery, seismic sensors, LIDAR data for autonomous vehicles. Whereas traditionally over the last decade compute and storage has become increasingly centralized in the cloud. This dichotomy prevents low--latency decisions and actions triggered by the data as well as incurs high cost of traffic. Edge computing provides a solution to this problem by making compute resources available close to the data sources. However, to make edge computing effective in practice, efficient deployment, management, and re--configurability of heterogeneous edge resources e.g. CPU, GPU, FPGAs within an end to end infrastructure such as in High Performance Computing (HPC) scenarios, is essential. We will define and develop an edge computing unit incorporating FPGAs as the re--configurable compute resource across different edge applications, along with software microservices as a resource broker that makes the unit easy to deploy into an end to end computing infrastructure and re--configurable and shareable by multiple applications. This proposed research by our company, in collaboration with a university, will prove feasibility of the resource broker approach, in which the user (e.g. scientists) can use the FPGA--based edge computing unit to solve edge problems such as in autonomous driving or flooding predictions. The co--operation between the FPGA--based edge hardware and the software microservices will enable efficient deployment, sharing, re--configuration, and management of the unit across data sources, users and applications. Once the feasibility of this approach is proven in Phase 1, we will proceed with productizing the edge computing unit (as one stop shop for edge hardware and software). The work described in this project will accelerate the deployment and adoption of edge computing in any scenario where low latency, secure, and high performance processing of major data sources, in environments with or without high bandwidth cloud connectivity, is important. This includes markets in the Commercial sector such as Smart Retail and Autonomous driving, Industrial sector such as mining and manufacturing, as well as government initiatives such as Smart Cities and disaster management. One example of such applications in disaster management: real--time decisions related to weather data in order to assess flooding or wind damage risk at a neighborhood granularity as a hurricane or thunderstorm approaches. Another example application is for at--home care of elderly with quick decisions to contact emergency services in the case of an at--home emergency such as a fall.

EdgeSense : Smart Sensors and FPGA based Heterogeneous Edge Computing Unit—CIRRUS360, 3017 Brookvale Drive, Richardson, TX 75082

Sudipta Sen, Principal Investigator, sudipta@cirrus3sixty.com

Sudipta Sen, Business Official, sudipta@cirrus3sixty.com

Amount: $213,255

Data influx from sources at the edge of the network is exploding e.g. from sensors to collect weather data, satellite imagery, seismic sensors, LIDAR data for autonomous vehicles. Whereas traditionally over the last decade compute and storage has become increasingly centralized in the cloud. This dichotomy prevents low--latency decisions and actions triggered by the data as well as incurs high cost of traffic. Edge computing provides a solution to this problem by making compute resources available close to the data sources. However, to make edge computing effective in practice, efficient deployment, management, and re--configurability of heterogeneous edge resources e.g. CPU, GPU, FPGAs within an end to end infrastructure such as in High Performance Computing (HPC) scenarios, is essential. We will define and develop an edge computing unit incorporating FPGAs as the re--configurable compute resource across different edge applications, along with software microservices as a resource broker that makes the unit easy to deploy into an end to end computing infrastructure and re--configurable and shareable by multiple applications. This proposed research by our company, in collaboration with a university, will prove feasibility of the resource broker approach, in which the user (e.g. scientists) can use the FPGA--based edge computing unit to solve edge problems such as in autonomous driving or flooding predictions. The co--operation between the FPGA--based edge hardware and the software microservices will enable efficient deployment, sharing, re--configuration, and management of the unit across data sources, users and applications. Once the feasibility of this approach is proven in Phase 1, we will proceed with productizing the edge computing unit (as one stop shop for edge hardware and software). The work described in this project will accelerate the deployment and adoption of edge computing in any scenario where low latency, secure, and high performance processing of major data sources, in environments with or without high bandwidth cloud connectivity, is important. This includes markets in the Commercial sector such as Smart Retail and Autonomous driving, Industrial sector such as mining and manufacturing, as well as government initiatives such as Smart Cities and disaster management. One example of such applications in disaster management: real--time decisions related to weather data in order to assess flooding or wind damage risk at a neighborhood granularity as a hurricane or thunderstorm approaches. Another example application is for at--home care of elderly with quick decisions to contact emergency services in the case of an at--home emergency such as a fall.