RINGS: Enabling Joint Sensing, Communication, and Multi-tenant Edge AI for Cooperative Perception Systems

San Jose State University Foundation, San Jose CA

Investigators

Abstract

Nationally, 40 percent of all traffic crashes involve intersections, with a quarter of all traffic deaths and about half of all injuries occurring in such locations. Recent advancements in vehicle communication and perception systems have led to significant improvement of situational awareness in modern connected and semi-autonomous vehicles. However, these light-based perception systems are limited by view blockage in busy traffic intersections and thus cannot reliably detect all objects. To improve sensing reliability, many autonomous systems rely heavily on multiple active sensors including 3D LiDAR, Radar and Ultrasound sensors. However, the improved perception reliability comes at the cost of bigger and more complex systems with multiple vendor-specific components, which cause much higher power consumption and increased maintenance, security and reliability issues. Moreover, there are no coordination or multiple access control mechanisms for nearby active sensors, thus potentially making nearby perception systems interfere with each other in busy traffic intersections. The goal of this project is to design a next generation system with multi-modal sensing and low-latency communication capabilities that share the same essential hardware including the spectrum band, baseband processing, and computing units. This design will result in a low-cost and compact perception device that provides cooperative sensing with nearby devices with improved resilience. To improve the device utilization and harness the power of multiple artificial intelligence (AI) services, this project proposes an edge computing framework to enable multi-tenant AI capabilities on the proposed perception system. The Federal Highway Administration (FHA) reports approximately 2.5 million traffic intersection accidents every year. The successful conclusion of this project will enable future sensing and communication systems to cooperate with the infrastructure and nearby devices automatically, detecting objects and mapping busy traffic intersections reliably, thus decreasing the number of accidents and fatalities. The proposed joint sensing and communication system will serve as a well-rounded, reference design for the development community to use in future connected and sensing systems. The proposed cooperative sensing mechanism will help lower the cost of individual devices while at the same time achieving higher resilience and minimizing interference between nearby systems. The proposed research will also lower barriers for undergraduate and K-12 students (especially underrepresented groups) to learn intelligent embedded systems design, internet-of-things, and AI systems while accessing state-of-the-art edge-computing lab system settings with remote testing, deployment, monitoring, orchestration, and over-the-air programming capabilities. The lessons learned and system developed in this project will serve as a plug-and-play resilient next-generation networking solution to unlock more innovative future 5G/6G projects and ideas. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.