Communication for Automated Vehicle Efficiency and Safety Systems

Abstract:

Overview: The human cost of transportation in the US reaches an unnecessary 33000 fatalities and 2.2 million injuries every year. The inefficiency of road transportation also burdens our economy, with only the cost of fuel wasted in traffic reaching $88 billion a year. To reduce this unnecessary human and economic cost, this proposal aims at advancing the concept of Cooperative Vehicle Efficiency and Safety (CVES) systems, through developing an information networking methodology for real-time dynamical system knowledge propagation. A CVES system uses advanced communication technology to coordinate the control and sensing actions of vehicles; this will enable cooperative (semi-) automated crash avoidance and vehicle drivetrain control. A critical need of CVES is reliable real-time awareness of the state of other coordinated vehicles. This task is particularly difficult in large scale deployments, as was recently shown for prototype cooperative active safety systems. In these systems, application and communication components become tightly coupled, leading to unreliability of the system when either component is strained. This proposal addresses the critical need for providing control oriented situational awareness for cooperative vehicles by proposing a model- and context-aware multi-resolution information networking methodology. The objective is to develop the concepts of model communication and its derived multi-resolution networking, to enable scalable propagation of CVES control and action related information (e.g., in stochastic hybrid system forms) instead of raw measurement data. Such methodologies will enable robust operation of CVES applications under both network and application behavior uncertainties. The knowledge generated through this research will be directly integrated with an educational program on Cyber-Physical Systems. The career plan proposed here is in synergy with PI's work in EcoCAR3 project.

Intellectual Merit: The concept of scalable modeled knowledge propagation will be realized through two specific ideas of 1) context-aware model communication, 2) scalable multi-resolution information networking. The concept of Context-aware model-communication relies on transmission of models (structures and parameters), instead of raw measurements. This allows for high fidelity synchronization of dynamical models of CVES applications over networks. One realization is to use stochastic hybrid systems models, with structures and parameters continuously evaluated, and provide awareness of the surrounding of a vehicle in the form of many concurrent dynamical models. Model communication can be coupled with multi-resolution networking concept through scalable representations of models and maps. Multi-resolution networking is achieved by in-network processing of scalable modeled information to adapt the fidelity of models to available network resources. This allows for adaptive knowledge propagation through unreliable networks, rather than simple data transfer. The result is robustness of CVES to network service variability. The proposed methodologies allow for a framework for concurrent composition of system component models. An expected outcome is a method for composition of unreliably synchronized stochastic hybrid systems, in particular derived for CVES.

Broader Impacts: The successful deployment of CVES, even partially, will provide significant societal benefits through reduced traffic accidents and improved efficiency; even a 5% reduction in accidents means saving hundreds of lives every year. The education component of this career plan is directed towards addressing the national need for CPS education, and will be coordinated with PI's efforts in the national educational EcoCAR3 competition. In addition to training graduate and undergraduate researchers in the interdisciplinary subject of CVES, three other specific activities are proposed. First, undergraduate students will be engaged through demos and a 5-year Capstone design project which will share components with PI's EcoCAR3 Capstone design course. Second, an educational package will be developed using the simulation tools that result from the proposed research, along with tutorials for hands-on training. The PI will also use research results to enrich his new graduate course on CPS, in consultation with WVU Education department (already collaborating in EcoCAR3). The outreach efforts will extend to K-12 students and public through demos at WVU summer camp, IEEE Women in Engineering demos, and in particular through EcoCAR3 events and social media campaign.