The Future of Driving: Control Barrier Functions and the Internet of Vehicles

The National Highway Traffic and Safety Association reports that 94% of serious car crashes are due to human error. Christos Cassandras, Boston University Distinguished Professor of Electrical & Computer Engineering, Head of the Division of Systems Engineering, and a co-founder of the Center for Information & Systems Engineering (CISE), has made monumental contributions to the research of network of vehicles and using systems to eliminate human error on the roads. 

Photo by IGphotography/iStock

Cassandras specializes in the areas of discrete event and hybrid systems, cooperative control, stochastic optimization, distributed optimization in network systems, and computer simulation, with applications to computer and sensor networks, manufacturing systems, and transportation systems. His recent research featured in Automatica on ScienceDirect, Optimal control of connected automated vehicles with event/self-triggered control barrier functions, highlights how the use of Control Barrier Functions (CBF) enables Connected and Automated Vehicles (CAVs) in traffic network conflict areas, therefore limiting the number of human-error crashes. 

​​Car crashes, whether caused by speeding, distractions or fatigue, are attributed to driver negligence or recklessness.“If we can automate at least a fraction of vehicles and reduce this 94% to under 50%, that would be quite something,” Cassandras said. “If you replace human drivers with computers, computers don’t sleep, drink, or blink. They love data, whereas [humans] become overwhelmed by data.” The CBF takes the wheel out of the driver’s hands and interacts with other vehicles artificially, removing human contact on the roads. 

Basic concepts of game theory and cooperation come into play when driving, according to Cassandras. Game theory shows that decisions between two or more decision-makers, in this case, drivers, show interdependence between the other’s decisions and cooperation, or lack thereof. Drivers primarily compete with one another but occasionally cooperate, letting another driver merge, pass, or go in front of them. 

“We selfishly try to find what’s best for us to get to our destination, to change lanes, to go through that intersection, and so on,” Cassandras said. With this new CAV technology, that competition would be erased and a “social optimum” would be created. The competition of who gets to go through an intersection or gets to merge first would no longer be in the hands of the driver. The decision of what is best would now belong to the computer in the car. 

For Cassandras, the keyword is safety. “Nobody will buy your automated car or any other expensive, elaborate piece of technology unless there are some guarantees that it’s going to behave safely,” he said. While achieving complete passenger safety and keeping track of what is surrounding the vehicle may be challenging, it is still possible. 

“If my car is behind yours, I don’t want to move too close to you and risk hitting you. That’s fairly easy to define and specify,” Cassandras said. “It is extremely hard to ensure this at all times, not just now and for the next two minutes, but forever.” The solution to this problem comes in the CBF. These functions are the constraints the vehicle operates under, including how close the vehicle can get to the vehicle in front or how fast the vehicle can move. “The [CBF] approach creates a mapping and transformation that says, ‘This is what you want. This is how I can achieve it,’” he said. These functions provide the guidelines for how these constraints can be complied with on the road. 

But, with these functions comes a sense of conservativeness. Cassandras said, “a perfect transportation system could be guaranteed if all vehicles drive five miles per hour, but that is clearly not what we want.” Therefore, a tradeoff arises between safety and speed with the CBFs. “I want to guarantee safety but at the same time I want to drive as fast as possible. That’s the state of the art. I’d like to be as efficient as possible without ever violating safety, but not at the price of being super conservative.” 

Finding this equilibrium between speed and safety is difficult to find but by using on-board calculations, it is possible. “You want all of these calculations begun on board in fractions of a second because that’s typically how much time there is to decide what to do next.”

There are two schools of thought moving forward: pro or anti Internet of Vehicles (IoV). Similar to the internet cell phone use, the internet of vehicles would be used for vehicles to communicate with each other on the road. Conversely, vehicle companies may not want to be a part of that “internet” and choose to use their own technology instead. 

Companies may want to equip their cars with their own sensors to communicate within their network instead of the broader IoV community raising larger issues related to weather or navigating obstacles. The IoV’s competitive advantage is left up to the consumers who decide whether or not to participate, creating a combination of autonomous and non-autonomous drivers on the road, “mixed traffic” as Cassandras referred to it.

Using a group of autonomous vehicles, a careless, distracted, or uncooperative driver can be mitigated—using a framework called ‘games with coalitions’. “It goes into the whole idea of how we can have cooperation among agents, autonomous devices, and still get things done safely and efficiently without having to worry too much about it,” Cassandras says. “If vehicles cooperate, they can sometimes neutralize the uncooperativeness of the human driver.”

Cassandras’ current and future grant projects include work with the National Science Foundation about delivering an online learning framework aimed at distributing travel demand in a given transportation network resulting in a socially-optimal mobility system that travelers are willing to accept. Additionally, Cassandras is working with Honda to explore and quantify benefits of using smart vehicles in highway traffic to improve flow currently sometimes subjected to high-variance unpredictable behavior causing ‘phantom traffic jams’.