“Automated vehicle (AV) technology will transform many aspects of mobility systems,” CEE Interim Chair and Prof. Yafeng Yin said. Prof. Yin provided an in-depth perspective on Automation, which is one of CEE’s five Strategic Directions. He recently participated in an interview that explored the many technical, philosophical, economic and emotional components that affect Automation, from its implementation to its growth and acceptance.
Q: What role does CEE play in fostering the growth of Automation?
A: “I can speak more on driving automation, since this is one of my research interests. CEE faculty also work on other automation applications such as automating the use of valves and gates to control the retention and flow of water through urban areas to prevent flooding.” he said. “At CEE, we develop enabling technologies for driving automation, e.g., cooperative sensing, path planning algorithms, and the AV safety testing. We also try to understand individual adoption attitudes towards the AV technology, and investigate public policies to guide its development and deployment. For example, we examine safety standards that AVs need to meet before they can be sold to the public, answering questions like ’How safe is safe enough?’ In addition, we spend a lot of time on establishing methodological foundations for planning and operations of mobility systems with mixed autonomy.”
Prof. Yin noted that automation is a complex topic that merges technical issues with social science, public policy and a variety of “soft” disciplines.
“For example, how do we ensure environmental justice in AV pilot deployments? We have witnessed more and more AV pilot deployments around the nation. Most of them are not deployed in neighborhoods where EJ communities may be concentrated in, with less well-maintained infrastructure and higher levels of pedestrian activity. This is understandable as it is much more challenging for AVs to navigate in those neighborhoods safely. A severe accident may cause a serious setback in the AV deployment and diffusion process. On the other hand, it has been reported that currently AVs struggle to recognize certain road user groups such as dark-skin pedestrians. The importance of developing unbiased AVs cannot be overstated. The remedy should start with simply increasing the number of images of dark-skinned pedestrians in the data sets used to train the systems. This thus has created a difficult dilemma that government agencies would face when planning AV pilot deployments. On one hand, allowing AV pilot tests in EJ communities would generate more knowledge to further improve the AV safety in the long term. On the other hand, in the short term, such tests impose higher risks to EJ communities. Some can even argue that this is against environmental justice, as the risk is unproportionally higher. Therefore, we are developing a quantitative economics model to derive policy insights that would guide governmental agencies to navigate this dilemma by making an explicit tradeoff between short-term risks and long-term gains.”
Q: Can you provide some recent background history on the development of Driving Automation and how it has evolved?
A: “More than 30 years ago, the concept was AHS–the Automated Highway System,” Prof. Yin said. “That’s more infrastructure-centric. The thought was to put a lot of sensors on the infrastructure, providing support for driving automation. The problem, however, was chicken and egg. In order to justify this infrastructure investment, you have to have vehicles using it. Without automated highways, few would buy those AVs. That’s why it failed. There was no business model for the AHS concept.”
“The current approach for driving automation is vehicle-centric. The idea is to put all the intelligence on the vehicle, and thus AVs do not have to rely on the changes in infrastructure to navigate a road network. This approach is working better, because carmakers can always make a pitch to certain segments of the population. In other words, there are viable business models.”
“The problem for such a vehicle-centric approach is that even with high-end sensors, individual vehicles will still have a limit in terms of their sensing capabilities. More importantly, from a societal perspective, it may not make economic sense to put all of the sensors on the vehicle side. In addition to vehicle-vehicle cooperation, it may be more cost effective to deploy roadside sensors at strategic locations to enable vehicle-infrastructure cooperation.”
“One of my students, Daniel Vignon, just defended his dissertation on economic analysis of infrastructure-assisted driving automation. His analysis shows that, from a societal perspective, there should be some sensors on the vehicle side and some sensors on the infrastructure side. Every dollar should bring the same bang for the buck–that’s the optimal split. This implies there will be different levels of vehicle automation that customers can purchase. There will also be different levels of infrastructure digitalization. Busy, more dangerous and heavily traveled areas, such as Intersections and interstates, will have heavy deployment of sensors.”
“Why is this not happening on a broad scale? There is a lack of funding at state/federal governments who own the infrastructure, and a lack of coordination with carmakers. So we are asking, ‘Can we leverage private/public partnerships to equip our roads?’ CEE faculty Peter Peter Adriaens and I have been working on a research project that aims to identify business opportunities/use cases and develop innovative financing models for scaling infrastructure support for driving automation.”
Q: Both “Autonomy” and “Automation” were used to represent one of CEE’s Strategic Directions. Is there a distinction between the labels?
A: “Yes, many people use these two terms interchangeably. However, to me, their implications are different. Automation refers to using machines to replace humans to perform a certain task. Perception, planning, maneuvering–these are different driving tasks. You have different levels of automation, depending on what’s being taken over by the machine. Level 5 is fully automated, where we use machines to completely replace all driving tasks.
“On the other hand, autonomy implies self-governance. The question is–’How do we interpret self-governance in automated mobility systems?’ To me, autonomy is the end game, the ultimate goal of automation. Can we achieve fully autonomous mobility? Look at the concept of human agency regarding their travels. The relationship between vehicles and human agency is complicated.”
Prof. Yin traced the social science that is woven into people’s complex relationships with the automobile. “In the past, we bought a vehicle and got to decide where to go, when to go, which route to take, and we had a sense of control, comfort and freedom. That’s why people purchased a vehicle. It’s not just about driving. It’s a status symbol; it offers a sense of control, comfort, and even shelter. A car comes with a lot of functionality and symbolism. When we own automated vehicles–we give up a certain level of agency–vehicle maneuvers such as car following and lane changes etc., are no longer controlled directly by us. We give these agencies up by our own choice. The question here is will we feel comfortable enough to forgo the agency on the choice of destination, route, or departure and arrival time? In this sense, a vehicle can be fully automated, but not fully autonomous.”
Q: Specifically, how does human agency play a role in automating traffic systems?
A: “From a system perspective, we’re going to leverage different levels of agency that people give up to better control our traffic systems. For traffic control, we have been using traffic rules, signs and traffic lights etc. Rules and signs are essentially static controls. For control actuators, we really have traffic lights (and maybe dynamic message signs). We use traffic lights to control arterials and freeways (ramp metering). Traffic systems are not fully controllable, as we don’t have many actuators.
“In the future, as people give up their agency, give up speed choice, and route choice, at that point, AVs can become the actuators. We may only need to control 5 percent of AVs to regulate or smooth a traffic stream. If 20 percent of vehicles are following our routing guidance, we can affect traffic congestion and reduce it. We can leverage the agency of route choice that 20 percent of vehicles give up to better distribute traffic demand spatially.
“Traditional route guidance systems, such as Wayze, cannot ensure compliance. If our objective is to control 20 percent of traffic, it’s a hard way to achieve it–We can’t mandate that people will follow the suggested routes, so it’s not very efficient due to the level of noncompliance. AVs, on the other hand, are controllable, if their owners give up their agency. With an AV, we can specify ‘This AV is my AV.’, and still dictate what the AV does. When we yield control to the system, that’s when we give up more agency. Everyone has a value associated with their agency. The value of agency varies from person to person. A mechanism, auction or negotiation, can be in place to incentivize AV owners to give up their agency. For example, the system identifies the first AV in a queue, and then asks, ‘if I pay you $5, are you willing to allow me to control your speed?’
“If the system controls the first vehicle, it can influence everyone afterward. The second vehicle gets less offered because the system will get less control. The system can estimate the benefit it will receive from controlling an AV to determine the maximum incentive it can offer. However, for some, you don’t even have to offer money to incentivize. They either don’t care about their agency or because they want to do something for the common good. Others may view cars as a source of freedom and are thus less likely to give that up. Age can be one of the affecting factors in the value of agency. So are gender and culture. That’s why the behavior component always complicates automating traffic. Transportation is a perfect area for people-first engineering. We start with people because people want to move from point A to point B. We design a system to satisfy people’s mobility needs. Understanding why people travel and how people travel has been a focus in traditional transportation engineering. An emerging research need is to understand attitudes and preferences of travelers towards various levels of agencies when interacting with AVs.”
Q: How is Automation affecting the future employment opportunities of students entering Civil & Environmental Engineering?
A: “We have to ask how and why are we getting into the domain of developing these types of technologies? Mechanical Engineering has traditionally designed vehicle hardware. Civil Engineering traditionally designs infrastructure. Transportation engineering studies how to manage the system when rubber meets the road. Moving into the future, vehicles become essentially software on wheels, software for perception, planning and control. For better planning and control, one has to anticipate the behaviors of the other participants in the traffic system. There is an interaction among each moving part. That’s where our transportation students have an edge, because we have been studying various participants in a traffic system, and model their interactions. That’s probably the reason why more students who graduated from our transportation program work in tech companies that develop AV technology or emerging mobility services. These job opportunities, typically with high pay, were not available a few years ago.
“Beyond transportation, our nation’s civil infrastructure is in the midst of a dramatic transformation. Infrastructure sensing technologies have grown ubiquitous, and actuation has also become more common. Such technologies enable the usage of various infrastructure to be controlled remotely and automatically, for the purpose of enhancing functionality and efficiency, improving safety and security, or to promote equity and fairness. This is the wave of the future for Civil & Environmental Engineering and it’s important to note as we encourage students to explore our field. There are such diverse opportunities available to them, from the traditional to the cutting-edge.”