By 2022, automakers have promised that all new cars and trucks sold in the U.S. will be able to look down the road, see if a vehicle is about to hit something, and apply the brakes if the driver fails to heed warning lights and buzzers. That industry-wide agreement, in lieu of a regulatory mandate, moves advanced driver assistance systems (ADAS) from the niche world of luxury vehicles to the mass market, and it will lead to a boom in sensor systems and controls in future vehicles.
Officials at companies that design and manufacture the sensors and electrical systems of vehicles say the growth of ADAS provides an opportunity to make vehicles safer and more capable, but adding more sensors is going to force changes in the supply chain, require designers to get creative on where to put all of the new systems, and give electrical engineers new challenges when dealing with new power- and data-transmission requirements.
“Something that was really an exclusive technology, just for import luxury vehicles only 10 years ago, is now becoming relatively common, all the way down to compact and subcompact vehicles,” says Andy Whydell, director of product planning for vehicle systems and functions at automotive supplier ZF TRW. “We need to be able to ramp up our production processes to accommodate that.”
While adding multiple sensors to 17 million vehicles per year by 2022 is a huge increase from the current rate of ADAS, Whydell says the industry-wide approach is already making the upcoming task easier. Vehicle producers have begun altering their processes, and suppliers are adapting quickly to the new demands.
“If they were only fitting a few radars a day, going through a manual alignment process is something that could be done off the production line at the end of assembly,” Whydell says. “But if every car is getting a radar, the alignment process has to be inline and automated.”
On the supply side, companies such as ZF TRW are dedicating entire plants to ADAS sensors instead of portions of facilities. Whydell adds that with demand growing so quickly, the supplier is planning on regional production in Europe, Asia, and North America to feed customer plants in those regions instead of centralizing production to one or two plants.
Aaron Jefferson, director of product planning and strategy for ADAS electronics at ZF TRW, adds that chipmakers are catering to automotive suppliers by developing processors and systems specifically for the market instead of offering off-the-shelf signal-processing solutions.
“In the past five or six years, sensors and processors have become more affordable. While the processing power has gone up, the actual component pricing has come down, so the capabilities continue to increase,” Jefferson says. As with other computerized components, next-generation sensors and processors are often smaller than their predecessors.
So, the systems needed to enable ADAS are getting smaller, less expensive, and more powerful. And moving from off-the-shelf parts to tailored, high-volume solutions should further lower costs – perfect for increasing ADAS use. But there are some significant challenges, says Nick Smith, business development director at Mentor Graphics, a company that makes software tools used to design automotive electrical systems and wiring harnesses.
Until very recently, each ADAS system – camera-based lane-departure warning alerts, blind-spot warning alerts, adaptive cruise control, rear-object detection – has been developed and marketed as a separate system. So each function tends to come with its own sensors, wiring requirements, and computer processors. With each new feature, car companies must purchase chips and figure out how to integrate multiple data streams.
“Each function required its own electronic control unit (ECU), so if an original equipment manufacturer (OEM) wanted to add a new system, that typically meant adding one or two additional ECUs. That trend is becoming cost prohibitive, and it’s created a packaging problem,” Smith says. “It’s becoming unsustainable to go from one ECU to 10, 50, or 150 processors on the car. The growth in ECUs has to be addressed at some point.”
That chip explosion is leading to two big problems – cost and packaging. While the cost of each chip may be falling, the number of chips is growing faster. Because of this, figuring out where to put all of those processors in a way that avoids heat buildup and other electronics concerns has companies doing weird things, such as putting an ECU in the rear spoiler and running power and data cables to it.
Centralized or distributed?
For car manufacturers and Tier 1 suppliers that create holistic systems that integrate multiple sub-safety systems, the answer is to consolidate processing. Instead of having radar, sonar, laser, and camera systems collect and process information at the data collection point, future systems will collect the data and send them to a few powerful, centralized processors. Those central computers will combine multiple data streams to create a real-time, three-dimensional view of the vehicle’s surroundings.
As Whydell puts it, “We’re trending away from independent systems to a central controller that’s receiving input from multiple sensors around a vehicle – more like a hub-and-spoke arrangement. All of those sensors feed their data into one location where a central ECU combines it to build a really good picture of what’s happening around the vehicle.”
With that hub-and-spoke system, sensors won’t need as much processing power, so they should get smaller and lighter. Those centralized processors will be more expensive, but eliminating all of the distributed chips currently on cars should lower the overall system cost. That’s the theory at least.
Smith says some early experiments with a centralized computer failed to produce the cost savings expected. One manufacturer, he says, tried the small-scale task of consolidating all of the electrical controls in its doors to a single chip – using centralized processing for the power windows, mirror controls, door locks, and safety sensors. The cost of the chips fell, but extra cabling to run data to the central computer added cost and weight. Smith says the overall cost and weight of the system were higher than a more-distributed version.
“Most cars have tens of kilos of copper wire on board. It’s expensive and bulky, and that cost has to be considered when you talk about centralizing computing,” Smith says. “Data transmission, physical harness layouts, and power use all have to be optimized together. How the architectures play out – in terms of consolidation of computing power versus distribution – is not so straightforward.”In addition to the weight and cost of wiring, other challenges to centralized computing include:
- Bandwidth – As the number of sensors on cars increase and the quality of their signals grow, getting all of that data to a centralized computer requires new, high-speed communications protocols, including Ethernet.
- Signal loss – Data from a car’s rear bumper may have to travel 30ft to 40ft to a central processor as cables snake around vehicles. Standard cabling systems, such as low-voltage differential signaling (LVDS) manage long distance and power at low weight, but long transmission distances act against low-cost solutions.
- Redundancy – Given that ADAS sensors are critical to safety, designers will likely have to follow the aerospace industry’s standards in creating backup processing units in separate areas of the vehicle – adding cost and requiring cable connections to two locations.
Jefferson and Whydell agree that centralization carries extra cabling complexity and cost, but they say that some consolidation will have to occur because managing dozens of independent safety systems is more problematic.
Finding the balance
Smith says optimizing weight when incorporating ADAS into vehicles is going to require a lot of trial and error, and sophisticated software modeling tools, as companies figure out which functions need to be centralized and which ones are better served by doing their computing at the point of data collection.
An example, he says, is power windows. They use a simple, on-off switch, but they need some intelligence to sense when there’s an obstruction in the window’s path so they don’t cut off peoples’ fingers. Centralizing such as simple function would add cost and weight and offer virtually no benefit, he says.
“You have components that by their nature are localized. Window motors don’t need to be part of a central data hub, nor do dome lights,” Smith adds. “If you start designing a backbone data bus, you can quickly learn that the backbone has 1,000 ribs coming off of it. That level of complexity isn’t necessary.”
ZF-TRW’s Jefferson says some of the design challenges should get easier now that automakers have committed to using the technologies in all new cars and trucks.
“Early on, packaging these features was a bit of an afterthought. It was a matter of squeezing the technology into an existing vehicle. Now, ADAS are a critical part of the vehicle planning process, so there’s room made and designed for early on,” Jefferson explains. “It’s more planned into the vehicle architecture and modeled throughout the design process.”
Whydell adds that even with a centralized hub-and-spoke model, some processing will be done at the point of data collection. However, systems in the future will have one small processor on board instead of one to handle data collection and processing and a second to handle communications.
“Some pre-processing of the data will be done at the sensors because the bandwidth needs would be too high to send raw data through a vehicle bus to a central processor,” Whydell explains.
Jefferson adds that several chipmakers are designing simplified processors that can crunch sensor data at the collection point and send only critical information back to a centralized processor. Systems such as those, he says, will allow ADAS designs to go from 95% of the computing being done at the sensor level today to a hub-and-spoke system by 2022.
Features enabled by centralizing those functions will outweigh the cost of cabling – mainly because suppliers will be able to more tightly integrate safety systems, creating a much better picture of what’s happening around a car.
“Now, it’s not unusual to find a car with a forward-looking radar for adaptive cruise control, a separate forward-looking camera for lane-departure warning, and a separate blind-spot radar system. You could have four different sensors in three systems, completely independent of each other,” Whydell says. “That’s not the optimal solution.”
Mentor Graphics Corp.
About the author: Robert Schoenberger is the editor of TMV and can be reached at 216.393.0271 or email@example.com.