US Ignite: Focus Area 1: Predictable Wireless Networking and Collaborative 3D Reconstruction for Real-Time Augmented Vision

点击翻译按钮获取中文摘要

Eliminating the line-of-sight constraint of human vision and machine vision, the developed network systems foundations of predictable wireless networked and 3D reconstruction will enable 'see-through vision' which will transform the ways humans and engineered systems interact with environments and thus have far-reaching impact on domains such as road transportation, public safety, and disaster response. This project develops the network systems foundation for a vehicle equipped with sensors and an augmented reality display to indicate the presence of other nearby vehicles hidden by obstacles. In collaboration with Wayne State University (WSU) police and Ford Research and leveraging the WSU living lab and the OpenXC open-source platform for connected vehicles, the project will take an integrated approach to the research, deployment, and dissemination of the wireless network systems for see-through vision. This project proposes a cross-layer framework for addressing physical-domain uncertainties and the interdependencies between wireless networking and 3D reconstruction, and it develops novel algorithms for predictable wireless networking and real-time wireless networked 3D reconstruction. Using the developed network system, this project will develop a see-through vision application for human-driving. The wireless networked see-through vision system will be deployed in the WSU police patrol vehicles, and the project team will outreach to the Detroit and State of Michigan police as well as open-source communities for broad adoption and deployment of the see-through vision system.With the bold objective of eliminating the line-of-sight constraint of human vision and machine vision, this project addresses wireless networking and 3D reconstruction challenges in a holistic cross-layer framework. By integrating research investigation with systems development and deployment, this project will make the following significant contributions: 1) Effectively leveraging multi-scale physical structures of traffic flows, the multi-scale approach to resource management in vehicular wireless networks not only ensures predictable vehicular wireless networking, it also transforms fundamental challenges of vehicular networks to ones similar to those of mostly-immobile networks, thus enabling the exploration of fundamental, generically-applicable principles and mechanisms for predictable wireless networking; 2) the multi-scale approach to joint scheduling, channel assignment, power control, and rate control enables predictable control of per-packet transmission reliability in the presence of fast-varying network and environmental conditions such as wireless channel attenuation, internal and external interference, data traffic dynamics, and vehicle mobility; 3) the real-time scheduling algorithm enables controllable exploration of real-time capacity regions for system-level optimization; 4) the collaborative 3D reconstruction model integrates visual sensors in a divide-and-conquer fashion, and it enhances the capability of networked vision as well as its robustness to physical uncertainties; 5) the co-design of collaborative 3D reconstruction and wireless networking permits adaptive communication capacity allocation to optimize the quality of 3D reconstruction; 6) the attention-aware see-through vision application creates a new research field of vision augmentation by uniquely integrating computer vision and computer graphics research and by proposing a practical solution for displaying augmented 3D vision.