The automobile has been around since the late 1800s, yet only in the past several decades have electronics become an integral player in the automotive world. Nowadays, automotive companies are trying to compete not only with each other, but with technology and time. As technology has progressed, so has the capability to integrate silicon and sensors into an automobile, which has gotten to a point where it is affordable enough and electronics are small enough that numerous electronic control units (ECU) can be installed throughout the vehicle. So as technology progresses, vehicles will be expected to contain state-of-the-art electronic technology in the same vein as consumer electronics. Herein lies the challenge; it is difficult for ECUs based on CPUs and GPUs to keep pace with consumer electronics due to multi-year chip development cycles and rigorous automotive quality and reliability qualifications. FPGAs may play an important role to close this gap by providing cutting-edge performance and the flexibility for system architects to customize their designs with the flexible (programmable) FPGA fabric. As FPGAs become more efficient and affordable, the goal is in sight.
Michael Hendricks, Altera’s Automotive Product Line Manager, has been developing and researching advanced vehicle control systems for much of his career. During this time span, he has determined that it is necessary for automotive technology to advance at a similar rate as consumer electronics. He explains, “As consumers, we have become accustomed to the blistering pace of innovation set by major smartphone suppliers who are releasing new product models every 18 months. We have been conditioned to now expect these same levels of performance, features and high-resolution displays in our automobiles, yet it is quite common for people to own the same car for many years or even decades.” This presents a major challenge for automobile manufacturers – how can they keep their vehicle technology relevant for the lifetime of the product? Hendricks points out that while it may not be possible to upgrade the technology in older cars, as they were not designed with this capability in mind, today car makers are increasingly turning to programmable electronic system architectures in order to keep pace with technology innovations.
“Compare the performance of modern consumer electronics such as a smartphone with that of a car’s infotainment system,” Hendricks explains, “and you would expect the same level of performance out of your infotainment system that you have become accustomed to on your smartphone.” It is not going to be possible to continue to move forward with the production of automobiles without the possibility of electronic upgrades. Some companies have already begun refreshing their hardware with higher frequency. Hendricks mentions that, “companies like BMW used to release a new head unit roughly every five years, and now they are refreshing them every two years.”
The key factor in the design of FPGAs is that they are considered programmable logic. Software can obviously be updated in a CPU, but the silicon of a computer chip cannot. Mr. Hendricks explains, “With the advent of programmable logic, it would be feasible to do a software and hardware update. FPGAs can be reconfigured or reprogrammed for different functions an infinite number of times.” Many vehicles sold today have software controlling a lot of the functions needed for driving functionality. “Tesla’s Model S already supports wireless software upgrades via a 3G connection,” Hendricks explains.
It is already possible to update a car’s software in order to keep up with the current expectations in functionality; however, with programmable logic on FPGAs, this can be taken a step further. The logic behind the execution of code can be updated on demand, helping the silicon function more efficiently in certain areas. “Because of this capability,” Hendricks says, “FPGAs may help OEMs keep pace with the latest trends by having a configurable or a reprogrammable hardware architecture system.” So when someone purchases a new vehicle, the vehicle’s electronics-enabled features can last much longer as new software and hardware updates can be performed at the dealership and at specialized maintenance shops. These software updates can be applied to many different functions of the vehicle, as a vehicle can contain many different FPGAs as these devices become smaller and cheaper. “Of course, safety and security are paramount” explains Hendricks, “And Altera has been a pioneer in this space as the first FPGA supplier to achieve IEC 61508 certification for functional safety which forms the foundation of the automotive functional safety specification, ISO 26262.”
The infotainment centers and dashboards now contain video screens that update on demand. These units contain ECUs that help control their functionality. However, these are not the only areas that contain ECUs; functions like power steering, gas monitoring, data metering, power locks, and power windows all function off of an ECU in the vehicle. “The Modern Daimler S-Class has over 100 ECUs in their luxury-class model. It is quite costly to support that number of ECUs, each with its own dedicated microcontroller or microprocessor,” mentions Mr. Hendricks.
The engineers designing the car are looking to reduce the total number of ECUs as these ECUs drive up the overall cost of the vehicle. The ECUs also consume a lot of energy from the vehicle, which the vehicle could be using to be more efficient. Currently, as FPGAs continue to develop, it is smarter to combine electronics inside the vehicle. “There is a trend towards sensor fusion and centralized architectures – especially in ADAS applications. As it stands, there are challenges to implementing a pure CPU-based architecture.. We are starting to see a hybrid approach with CPUs and GPUs mixed with SoC FPGAs,” Hendricks explains.
According to ARM, vehicle compute performance is expected to increase by a factor of 100 over the next 9 years. “The beauty of the FPGA is that you have this vastly configurable and programmable silicon with a very wide parallel processing architecture that allows Tier 1 and OEM developers to continue to evolve their algorithms and make design tweaks all the way up until they near production release,” Hendricks stated. In addition to pre-production release, it will also be possible to upgrade the logic post-production. This allows very specific applications to be continually monitored and updated with the consumer’s needs.
There are several specific current applications that FPGAs are being designed into. One application that seems particularly futuristic is the virtual cockpit. “A virtual cockpit is being created inside the vehicle where analog gauges and dials are being converted for reconfigurable digital displays, such as in the instrument cluster,” Hendricks clarifies. Instead of speedometers and miles-per-gallon being displayed on the dashboard, these instruments will be displayed on a windscreen of sorts. Along with displaying pertinent information, there is a push to make these instruments more interactive. “We are starting to migrate away from your traditional touchscreen/capacitive touch and we are looking at new ways to control the HMI, such as gesture recognition and using cameras and time of flight sensors to recognize subtle hand gestures to control the electronics of the vehicles,” Hendricks projects.
관련 상품 참조
Low latency is key. For example, one particular hardware trend in modern vehicles is the proliferation of backup cameras with displays embedded in rearview mirrors and head unit consoles. When a car is put into reverse, the central information display projects a rearview camera view so drivers can see everything that is behind them as they back up. The requirement for these systems is that upon startup of the vehicle, the live video image captured from the rearview camera displays instantly allowing the driver to safely reverse their car. Hendricks addresses this need with Altera’s innovate MAX 10 family of FPGAs by describing, “FPGAs are intrinsically faster to startup; Altera’s Max 10 FPGA with integrated flash allows it to boot and display an image in the order of tens or hundreds of milliseconds.” CPUs, on the other hand, do not have the same capability for fast on (boot) time. “With traditional CPU architectures, the boot time is on the order of several seconds – too long for a driver to wait for the rearview camera to display an image on screen prior to reversing the vehicle,” Hendricks explains. FPGAs blow that startup time out of the water, making them a better solution to the camera startup time problem.
As we move towards the future of autonomous driving, it’s all about safety and quality of life. Forward-facing cameras in cars are now used for a range of applications, giving rise to a whole new set of complex processing. “We’re moving towards a heavy analytics-based architecture in ADAS through a combination of sensors including camera, radar and LIDAR. With these advanced sensors and vision analytics, features such as automatic emergency braking, traffic sign recognition, pedestrian detection and lane departure warning are all becoming a reality,” says Hendricks. Many of these functions are possible with the advent of convolutional neural networks (CNNs) – a type of feed-forward artificial neural network increasingly used for image and video recognition. When FPGAs are implemented into the ECUs of the vehicle it can use CNN algorithms to process images coming from the vehicle’s cameras.
“In ADAS vision applications, CNNs can be used for identifying objects around the vehicle and classifying them. For example, a CNN algorithm can determine if the object in front of the vehicle is a pedestrian, cyclist, animal, traffic sign or other object.,” Hendricks explains. The camera can then make an educated decision to tell the driver of the approaching object or even activate machine-control of the car to move out of the way in the case of an emergency. This can also be applied to traffic lights and signs, allowing the FPGA to make smart decisions and keep driver safe. The processing power required to run these complex algorithms can be significant and power consumption and heat dissipation are critical attributes for the small form factor ECUs with no ventilation or active cooling to dissipate the heat generated by the electronics. In a recent whitepaper published by Microsoft, Altera FPGAs were found to run at a fraction of the power of a GPU for comparable CNN data sets.
FPGAs are being implemented in place of as well as in conjunction with CPUs as a faster and more efficient solution to hundreds of ECUs inside a vehicle. FPGAs provide more performance-per-watt, scalability and customization capability. As Hendricks states, “FPGAs can help lower the total cost of ownership through integration, reduction of external components, faster time to market and a unified development flow. ” By implementing FPGAs, the number of ECUs can be decreased. Companies like Tesla and Audi are already implementing Altera’s Cyclone family of SoCs and FPGAs in their infotainment and ADAS systems to optimize communication connectivity and boost performance. It’s a bright future for programmable logic in automotive applications and we can expect to see greater proliferation of these advanced FPGA technologies being adopted in tomorrow’s advanced processing functions like self-driving cars, augmented reality HUDs and artificial intelligence processing.