Workshop Agenda
Following talks will be given within the 4-hour workshop session on Friday Oct. 30, 2020
The following agenda is a draft version, which will be finalized once all invited speakers are confirmed.
8:00-8:30 a.m. PDT Yongkang Liu
Sensor Fusion of Camera and Cloud Digital Twin Information for Intelligent Vehicles
With the rapid development of intelligent vehicles
and Advanced Driving Assistance Systems (ADAS), a mixed
level of human driver engagements is involved in the
transportation system. Visual guidance for drivers is essential
under this situation to prevent potential risks. To advance the
development of visual guidance systems, we introduce a novel
sensor fusion methodology, integrating camera image and
Digital Twin knowledge from the cloud. Target vehicle bounding
box is drawn and matched by combining results of object
detector running on ego vehicle and position information from
the cloud. The best matching result, with a 79.2% accuracy
under 0.7 Intersection over Union (IoU) threshold, is obtained
with depth image served as an additional feature source. Game
engine-based simulation results also reveal that the visual
guidance system could improve driving safety significantly
cooperate with the cloud Digital Twin system.
8:30-9:00 a.m. PDT Chris Schwarz
Digital Twins: From Macro to Micro
Simulation is now an integral part of automated vehicle testing programs, and many more miles are simulated than are
driven on the road. High fidelity digital twins are required for accurate simulations, and rely in turn on large amounts
of data collected from the real world, in the form of HD maps and sensor data. While simulation-based AV testing has been
used to massive effect for testing edge cases and large numbers of miles, it will be necessary for future simulations to
be even more realistic. They will have to enable AV simulations in different seasons of the year and under different
weather conditions (beyond current capabilities). They will have to include increasingly rural environments where lane
markings and surface conditions are poor. This will motivate the use of high-fidelity vehicle dynamics models that offer
realistic performance in extreme operating regimes. Sensor models and their validation will need to become more specific,
narrowing in on different brands and specifications. We summarize three studies at the National Advanced Driving Simulator,
both simulation and on-road. Based on this research, we discuss the benefits of a high-fidelity digital twin and its
application to the three study areas. While we begin with the more-common macro applications that involve many miles
and/or many vehicles, we end with the micro. There is much benefit to be gained from studying the interactions of one
driver with one vehicle, or between one AV and a vulnerable road user, in a realistic simulation environment
9:00-9:30 a.m. PDT Ruimin Ke
Design and Field Implementation of IoT Systems with Real-Time Traffic Video Analytics
Video analytics is driving a wide range of smart mobility applications with great potential to benefit transportation and infrastructure systems.
Most applications, e.g., traffic surveillance and autonomous driving, require not only high intelligence but also real-time operation.
However, real-time video analytics is considered the “killer app” for the Internet of Things (IoT) and often bottlenecked by the large
amounts of video data, high computational cost, and limited network bandwidth. In this talk, I will introduce the challenges and advances
in real-time traffic video analytics and their applications in transportation IoT. Specifically, I will talk about our recent work on the
design and field implementation of such IoT systems for smart infrastructure and automated vehicle testing.
9:30-10:00 a.m. PDT Yang Zheng
Integrating Autonomy into Traffic Systems: Scalable Control and Optimization
Traffic systems are undergoing a fundamental transformation. The emergence of connected and autonomous vehicles (CAVs)
provides tremendous opportunities for sustainability, efficiency, and safety. Before 100% CAVs are introduced in traffic
systems, there will be a long transition period of mixed mobility, where both human-driven vehicles and CAVs co-exist.
Timely and important questions are: can a small number of CAVs benefit urban mobility, and, if so, how? Moreover, the
design and operation of CAVs in traffic systems are becoming increasingly complex, due to system size, connectivity,
and heterogeneity. This talk presents contributions in distributed control and scalable optimization techniques, which
are critical for integrating autonomous vehicles into the existing traffic systems. In particular, I will provide
theoretical evidence of the high potential of CAVs in mixed mobility, and I will further show how to exploit sparsity
structures to design scalable and distributed algorithms that allow operating CAVs in large-scale mixed traffic systems.
10:00-10:30 a.m. PDT Prashant Tiwari
IoT and Smart Connected Vehicles Enabling MaaS Opportunities
The transportation industry currently is undergoing a huge transformation – estimated by some to be the biggest
transformation in its history – driven by Mobility as a Service (MaaS) opportunities and IoT revolution.
Vehicles increasingly are being equipped with embedded connectivity. By 2025, the Automotive Edge Computing
Consortium predicts that the connected car market will be worth $150 Billion, with 100 million connected
vehicles on the road. Analysts indicate that connected car capabilities already influence 6 out of 10 vehicle
purchases, giving vehicle manufacturers the ability to open new revenue streams, diversify core activities,
and improve profit margins by selling software and connected services. Smart vehicles are increasing becoming
most important verticals in IoT platform and processing the data from connected vehicles is one of the big
challenges all automakers are currently facing. But these challenges also provide opportunities to build hyper
connected cyber-physical systems (also called Digital Twin) with several interesting use cases like personalized
intelligent driving and smart traffic management.
10:30-11:00 a.m. PDT Daniel Work
Traffic Flow Smoothing at Scale
This talk will discuss modeling, simulation, and field demonstration advancements that are needed to control automated
vehicles to stabilize traffic flow at scale. Prior work on a closed course established that automated vehicles can
eliminate human-generated phantom traffic jams that seemingly occur without cause, reducing fuel consumption by up
to 40%. The talk will describe the research challenges and progress towards demonstrating traffic flow smoothing with
a fleet of connected and autonomous vehicles on the I-24 Smart Corridor in Tennessee, as part of the CIRCLES Consortium.
11:00-11:30 a.m. PDT Kanok Boriboonsomsin
Field Implementation of IoV Application for Improving Energy Efficiency of Heavy-Duty Vehicles
In recent years, the emergence of the Internet of Vehicles (IoV) technology has inspired numerous applications targeted
at improving existing transportation systems. As one of the widely studied applications of IoV technology, connected
eco-driving utilizes Signal Phase and Timing (SPaT) information from the upcoming traffic signal to design speed profile
for the connected vehicle to pass through the signalized intersection in an energy-efficient manner. To date, the majority
of connected eco-driving studies have been conducted through numerical or traffic simulations. Few studies have implemented
the application on real vehicles, and even fewer have been focused on heavy-duty trucks. In this presentation, the field
implementation of a connected eco-driving system for heavy-duty trucks that is enabled by cellular communications will be
described. Results and lessons learned from field trials of the system along two major freight corridors in the City of Carson,
California, will also be discussed.
11:30-12:00 noon PDT Jibonananda Sanyal
CTwin – The Chattanooga Mobility Digital Twin – Heralding the Next Generation in Observability and Control
Congestion continues to be a significant contributor to energy inefficiency in the US. In 2017, highway congestion
wasted over 3.3 billion gallons of fuel and caused 8.8 billion hours of lost productivity. Trucks accounted for $20
billion (11%) of the cost representing 7% of traffic. The talk focuses on the development of the City of Chattanooga’s
Digital Twin, named CTwin, that has delivered a regional-scale, open, scalable, real-time situational awareness platform
that brings in 500+ primary data streams from 7 proprietary vendor systems across 3 institutions, in addition to 40+
distinct data layers to create a capability that is unique for the Vehicle Technologies Office. Metrics based on industry
standards (MAP-21, Automated Traffic Signal Performance Measures - ATSPM) have been formulated to compute energy usage
estimates, regional speed, volume, and fuel used, as well as derive travel time reliability and percent stops on red,
besides others. Advanced data science approaches on real-time feeds have produced novel and exceptionally well received
data visualizations. Further, data-driven discoveries have led to an implementation of a control approach that is starting
to yield significant energy savings for the city. Talk will focus on the research, the developed capabilities, and the
anticipated outcomes of this work.
Workshop Organizers
