It is rare for an entire industry to experience a full-fledged revolution, and when these occurrences do in fact happen, they’re very far and between. According to NXP Semiconductor Vice President of Advanced Driver Assistance Systems Dr. Kamal Khouri, he believes the automotive industry is currently experiencing up to three revolutionary instances in connectivity, autonomy, and electrification. While speaking in a keynote event at NI Week 2018 in Austin, Texas, last month, Dr. Khouri stated the driving force behind these innovative revolutions derives from our responsibilities around the safety and security of autonomous vehicles and technologies, so when these products are delivered, they’ll work as expected.
“If you break down the problem of enabling self-driving cars, it can be categorized in three components—sensing, thinking, and acting,” says Dr. Khouri when he delivered his opening remarks during the keynote panel.
When it comes to autonomous vehicles, Dr. Khouri states sensing is largely about developing a model of the world around the car, and determining what the obstacles, free space, and agents are moving around the vehicle.
“These agents can be other vehicles or vulnerable road users like cyclists or pedestrians,” says Dr. Khouri at the keynote. “Once we develop this dynamic view around the vehicle, we have to decide what we want to do based on our judgment. This (our judgment) is critical and has to be correct because once that decision is made, you have to act.”
Dr. Khouri noted how the mechanisms that drive, steer, and stop a vehicle will require an immense amount of computing, which will be opposing itself in certain scenarios. This is where machine learning and artificial intelligence (AI) comes into play, namely when the vehicle is trying to determine its immediate surroundings. This is generally achieved by using probabilistic (meaning there is a chance you don’t get it right the first time) computing to see, detect, and classify an object.
While Dr. Khouri mentioned the driver still must make independent decisions that have to be correct the first time around, the AI of an autonomous vehicle (when in the same situations regarding decision-making), must also deploy deterministic methods to ensure their final decisions are correct and based on safety computing. The safety aspect of these systems, Dr. Khouri notes, gets even more complicated.
“In traditional automotive, we talk about (and are comfortable with) functional safety,” says Dr. Khouri. “You have an electric component or system and ensure it’s functioning correctly. If it fails or something goes wrong, it can detect and report that error somehow or correct the issue on its own if possible.”
Dr. Khouri noted how other safety dimensions are introduced in this “new realm of autonomous driving.” He cited behavioral safety, which covers policies a vehicle must follow to drive correctly, like following rules of the road, not driving on the curb, or stopping in the presence of a pedestrian (just to mention a few). Next, there is environmental safety regarding how vehicles react to changes in its environment, like intense traffic or driving through a huge puddle while it’s raining, and seeing if the sensors react properly when detecting the subsequent splash.
“The sensors detect something and need to determine in real time, if it’s a vulnerable road user or something the car can drive through,” Dr. Khouri notes. “These particular components would be connected to infrastructure, other vehicles, and within the vehicle itself. The network has become extremely complex with hundreds of thousands of code lines that must make these calculations to determine what course of action to take (usually) in under a second.”
When he touched on the security aspect of vehicular autonomy, Dr. Khouri covered it from a unique angle that is largely overlooked in this industry—cybersecurity. Dr. Khouri noted how vehicle hacks are on the rise, with 25 notable instances being documented, with the largest affecting 1.4 million vehicles and costing billions of dollars.
“Data being collected on you as a driver (or passenger), location of your vehicle, ability of pushing down services to the vehicle, sheer system complexity, and number of coding lines can make your car vulnerable to malicious code that can extract personal and card (among other kinds) information,” says Dr. Khouri when addressing why hacking in autonomous vehicles is such a concern. “By putting something like V2X infrastructure into place, every car becomes like a cellphone. The immense number of connected vehicles around the world will increase the vulnerability of hacking.”
Dr. Khouri went on to say it was the job of all people involved in the automotive industry to protect privacy and prevent unauthorized access, but also noted how the concerns aren’t solely focused on stealing personal data.
“Vehicles on wheels can be used as weapons, which we’ve seen happen in recent years; so how do we make sure malicious agents don’t get control of these vehicles and wreak havoc?” asks Dr. Khouri. “By doing so we will increase the safety of autonomous driving and potentially save lives with the introduction of this preventative technology.”
In his final remarks, Dr. Khouri noted how the automotive industry must promise to its customers that they will deliver great automotive solutions in the form of safety, security, and reliability, when it comes the future widespread development and deployment of autonomous car technologies.
“When we go to a dealership nowadays, we’re there with the aim of buying a car without worrying about taking it back or fixing problems and having to deal with issues,” Dr. Khouri states. “With autonomous technology slowly integrating in the automotive industry, this is a promise we must continue giving to our customers.”