Mohan M. Trivedi

Distinguished Professor of Electrical and Computer Engineering, University of California, San Diego, USA

Looking at Humans in the Age of Self Driving Vehicles

Abstract: With recent advances in embedded sensing, computing, machine perception, learning, planning and control, intelligent vehicle technology is moving tantalizingly closer to a future with large-scale deployment of self-driving automobiles on roadways. However, we are also realizing that many important issues need deeper examination so that the safety, reliability and robustness of these highly complex systems can be assured. Toward this end, we highlight research issues as they relate to the understanding of human agents interacting with the automated vehicle, who are either occupants of such vehicles, or who are in the near vicinity of the vehicles. Self-driving and highly automated vehicles are required to navigate smoothly while avoiding obstacles and understanding the high levels of scene and activity semantics. For achieving such goals, further developments in perception (e.g., drivable paths), 3D scene understanding, and policy planning are needed. Designing fully autonomous robotic vehicles that can drive on roads does typically did not require models of drivers and how they interact with vehicles. In contrast, design of intelligent driver assistance systems, especially those for active safety that prevent accidents, requires accurate understanding of human behavior, modeling of human-vehicle interactions, activities inside the cockpit, and prediction of human intent. A human-centered framework for a distributed intelligent system includes the driver, vehicle and environment as three key components. The main idea is to develop an approach to properly design, implement and evaluate methods and computational frameworks for distributed systems where intelligent robots and humans cohabit, with proper understanding of mutual goals, plans, intentions, risks and safety parameters. We emphasize the need and the implications of utilizing an holistic approach, where driving in a naturalistic context is observed over long periods to learn driving behaviors in order to predict intentions and interactivity patterns of all intelligent agents. Moving toward vehicles with higher autonomy opens new research avenues in dealing with learning, modeling, active control, perception of dynamic events, and novel architectures for distributed cognitive systems. This presentation will give examples of some of the accomplishments in the design of such systems and also highlight important research challenges yet to be overcome.

Biography: Mohan Trivedi is a Distinguished Professor of Electrical and Computer Engineering and founding director of the Computer Vision and Robotics Research Laboratory and LISA: Laboratory for Intelligent and Safe Automobiles at the University of California at San Diego. Trivedi has mentored over 30 PhD, 15 Post-Doc, 75 MS scholars and has presented over 80 keynote/plenary talks. Currently, Trivedi and his team are pursuing research in intelligent vehicles, machine perception, machine learning, human-robot interactivity, driver assistance, active safety and intelligent transportation systems. He received the IEEE ITS Society’s highest award “Outstanding Research Award” (2013) and LEAD Institution Award for LISA (2015). Trivedi’s team has played a key role in several major research collaborative initiatives. These include human-centered vehicle collision avoidance, lane change/turn/merge assistance systems, vision-based passenger protection system for smart air bags, predictive driver intent and activity analysis systems, and distributed video arrays for transportation and homeland security applications. He serves regularly as a consultant to industry and government agencies in the USA and abroad. He is a Fellow of IEEE, SPIE, and IAPR. You can find more about his work here.

Susana Sargento

Co-founder of Veniam, and Professor at the Department of Electronics, Telecommunications and Informatics, University of Aveiro, Portugal

The Role of Vehicular Networks in Intelligent Transportation Systems

Abstract: There has been a strong focus in the last 10 years on the development of vehicular network technology and mechanisms to build a reliable wireless mesh network to forward data packets in a multi-hop fashion between the vehicles and the Internet via road side units or access points. Although the vehicular technology, IEEE 802.11p, has been standardized some years ago, the number of real experiments and deployments are limited to the lab environment and small-scale trials. Moreover, the vehicular network mechanisms have been mostly tested in simulation environments, which are far from the reality of the network characteristics in real scenarios. Seeking to develop and test vehicular network mechanisms in real environments, and to explore the extent to which a vehicular mesh network can serve as a reliable and secure wireless infrastructure for mobile Internet access and the Internet of Things at street level, we built a city-scale testbed comprised of 404 connected public transit buses, 25 municipal service vehicles (garbage trucks) and 47 roadside units. This experimental platform is currently providing free WiFi to passengers in the city of Porto, Portugal. Moreover, these vehicles gather large amounts of data not just from their own sensors and the devices of the bus passengers, but also from other wireless devices that are spread around the city. This talk describes latest advances in vehicular network mechanisms and how real and large-scale vehicular network infrastructure is able to provide meaningful performance results, test new network mechanisms and develop new applications for both citizens, transportation and the city, in an Intelligent Transportation System perspective.

Biography: Susana Sargento is an Associate Professor at the University of Aveiro and a Senior Researcher at the Institute of Telecommunications, Portugal, where she is leading the Network Architectures and Protocols (NAP) group. In March 2012, Susana has co-founded a vehicular networking company, Veniam, which builds a seamless low-cost vehicle-based internet infrastructure. She has more than 15 years of experience in technical leadership in many national and international projects, and worked closely with telecom operators and OEMs. She has been involved in several FP7 projects (4WARD, Euro-NF, C-Cast, WIP, Daidalos, C-Mobile), EU Coordinated Support Action 2012-316296 “FUTURE-CITIES”, national projects, and CMU|Portugal projects (DRIVE-IN with Carnegie Melon University). She has organized and has served as TPC-Chair in several international conferences and workshops, such as ACM MobiCom 2009 Workshop CHANTS, IEEE Globecom and IEEE ICC. She has also been a reviewer of numerous international conferences and journals, such as IEEE Wireless Communications, IEEE Networks, IEEE Communications. Her main research interests are in the areas of self-organized networks, ad-hoc and vehicular mechanisms and protocols, such as routing, mobility, security and delay-tolerant mechanisms, resource management and virtualization in both network and cloud resources, and content distribution networks. Susana is the winner of the 2016 EU Prize for Women Innovators.

Antonio A. F. Loureiro

Professor of Computer Science, Federal University of Minas Gerais, Brazil

Vehicular Traces Are More Than Just Points on a Map

Abstract: Simulation is the most adopted approach to evaluate Vehicular Ad hoc Network (VANET) solutions. Furthermore, the reliability of these results depends on vehicular mobility models to represent the real network topology. Usually, simulation tools use mobility traces to build the network topology based on the existing contacts between vehicles. However, the traces’ quality, in terms of spatial and temporal granularity, is a key factor that impacts directly the network topology and, consequently, the evaluation results. In this talk, we show that existing real vehicular mobility traces present spatial and temporal gaps that affect their use to evaluate VANET solutions. We discuss a method to fill those gaps, leading to “calibrated” traces. This solution was applied to calibrate existing traces, and the results reveal that indeed the gaps may result in a network topology that differs from reality, leading to invalid results. We also present an upper bound in terms of communication capacity for original and calibrated traces.

Biography: Antonio A. F. Loureiro received his B.Sc. and M.Sc. degrees in Computer Science from the Federal University of Minas Gerais (UFMG), Brazil, and the Ph.D. degree in Computer Science from the University of British Columbia, Canada. Currently, he is a professor of Computer Science at UFMG, where he leads the research group in ubiquitous computing and ad hoc networks. His main research areas are ubiquitous computing, vehicular ad hoc networks, sensor networks and distributed algorithms. In the last 15 years he has published over 250 papers in international conferences and journals related to those areas, and also presented tutorials at international conferences. In December 2015, he was awarded the IEEE Communications Society Ad Hoc and Sensor Networks Technical Committee recognition award with the citation “for his contributions to the design, modeling and analysis of communication protocols for ad hoc networks”.