Automation has the potential to improve traffic safety and offer other benefits to drivers, pedestrians and society as a whole. The ability of sensor systems aboard automated vehicles (AVs) to collect and provide information that guides decision-making can reduce traffic accidents and optimize traffic flows.
CEE Associate Prof. Neda Masoud highlighted a project funded by the National Science Foundation (NSF) that focuses on Cyberphysical Systems. Prof. Masoud and CEE Prof. and Interim Chair Yafeng Yin are conducting research to leverage these automation advances.
“The idea is that connected AVs (CAVs) offer a lot of opportunities and anticipated benefits, but for these benefits to be realized, we have to devise systems to leverage these capabilities,” Prof. Masoud said, as she explained the concept of platooning. “A vehicle platoons is like a train of vehicles that are moving together with small gaps between them–like when people are tailgating–but in a safe manner because we have the sensor systems and connectivity between vehicles to ensure the smaller gaps do not make this risky. CAVs can move smarter and in a more coordinated fashion, and as a result will offer more energy efficiency, less traffic delay, andseveral environmental and mobility benefits.”
Platoon formation provides benefits because of these small gaps. “We are able to fit more vehicles on the road, so the capacity on roads increases, and fuel efficiency increases because the aerodynamic drag on platoon members is reduced.” Prof. Masoud said. She added that there are two main research areas the team examined:
1. The way in which a vehicle actually moves. How do vehicles decide where they should be next? How do they adjust speed and acceleration? “The conclusion is that typically, we use optimal control models, which are short-sighted because they are computationally intensive,” Prof. Masoud said. “They can only account for the immediate environment of the vehicles when trying to decide what the vehicle should do next, and are not accounting for the big picture, or network view. We proposed a way to combine local information alongside strategically condensed network-level information to leverage the platoons and the benefits they have to offer to the greatest extent possible.”
2. Stability of platoons–with multiple vehicles in a platoon, the degree of benefit each vehicle experiences will depend on where the vehicle is positioned in the platoon, as well as the characteristics of other vehicles in the platoon. “If you are behind a truck, you get more benefits than the first vehicle in the platoon by taking advantage of a larger reduction in the aerodynamic drag,” Prof. Masoud said. “If vehicles want to be selfish, they might find that it is in their best interest to keep switching platoons, thereby creating instability in the traffic stream. How do you form platoons that are behaviorally stable, where no one can benefit from switching between platoons?” Most of the work in this research area was computational and employed simulations to test at different levels of granularity.
Prof. Masoud explained that these topics focused on mostly operational decisions. “It’s done through a pricing mechanism,” she said. “When you join a platoon, you pay in exchange for the fuel efficiency benefits you get. Other platoon members that experience less fuel efficiency, but their presence is required for the formation of the platoon, get compensated for their participation. Pricing is done so no one will benefit from switching to a different platoon. If enacted in the real world, these pricing mechanisms would be implemented in an automated manner through bidding.”
Prof. Masoud noted that the best case scenario would be implementing platoons with fully driverless vehicles; that is, level 5 AVs,with no steering wheel. “You can have platoons at Levels 3 and 4 also–as long as we have connectivity. Connectivity is important so vehicles can be in constant communication. Some level of Automation is required, but the vehicle does not have to be completely driverless.”
Another project that Prof. Masoud highlighted focuses on cybersecurity of CAVs, and is supported by the Center for Connected and Automated Transportation (CCAT). CCAT is funded by the U.S. Department of Transportation.
“CAVs have more attack surfaces compared with traditional legacy vehicles,” Prof Masoud said. “There are a lot of sensors on board a CAV that could be hacked. The information that a CAV receives through the communication channel may also be manipulated.” Researchers are investigating ways to identify and detect attacks, as well as exclude and isolate sources of information that are compromised to keep others safe. “We propose new mathematical models that describe the interactions and the exchange of information between vehicles in a platoon,” Prof. Masoud said.
“We will need to have redundant information–not just one source of information. For example, I cannot only depend on a sole GPS device, in case it gets compromised. If that happens, and I don’t have other sources to recover the true information, there will be no path forward. We need diversified sources of information so they cannot be compromised in the same way. Then the question becomes, how do you retrieve the true signal, as different sources (e.g., sensors) may provide slightly different values? The team devises tools to make a system more secure so that one person cannot easily hack an entire system.”
A third project that Prof. Masoud highlighted is occurring with support from Denso International America and has a goal of enhancing safety at intersections. This research uses connected vehicle data to devise safety applications at intersections that target vehicles and pedestrians. “As a vehicle approaches an intersection, it sends information to a roadside unit and other vehicles,” Prof. Masoud said. The roadside unit collects information from all of the vehicles, which provides a big picture of the intersection. The same big picture can be secured by sensors deployed at the road side, such as cameras. “The road-side unit knows where vehicles are in relation to one another. It can use cameras to identify pedestrians and other objects. With its comprehensive view of the intersection, the unit can utilize our devised models and assign a risk level to different drivers. The intersection unit can then send alerts, warnings or messages to drivers, or even adjust traffic signals,” Prof. Masoud said.
This research is mostly based on data-mining methods and utilizes a pipeline of real-time data from intersections. This information then feeds into models, with the outcome being the risk level assigned to different drivers, and consequently the identification of risky scenarios.
The research only measures the trajectory of the vehicle, not information about the driver. “From the observed trajectory, you make a conclusion about what type of driver is behind the wheel without knowing anything about the person,” Prof. Masoud said. “Factors examined include speed, acceleration, vehicle jerk, angles, and other trajectory information. All of this is studied with respect to what’s happening in relation to other events and vehicles that are surrounding the vehicle source. If a vehicle is traveling at 40 mph without anything else around, that’s fine. With a person crossing the street in the path of that vehicle, that same speed is not safe anymore.” The roadside unit can detect that, send an alert, and help avoid an accident.