This article also appears in
Subscribe now »

TARDEC is using COTS components to simplify integration of sensors that pave the way to autonomous military mobility systems. (TARDEC image)

Military vehicles battle for autonomy at lower cost

The race to add digital capabilities to military vehicles is challenging engineers as they strive to create sophisticated systems that can be altered and upgraded without issues. Networking and architectural strategies rely on standards that help keep costs under control while giving warfighters more automated features and functions.

Many mainstream ground vehicles haven’t seen the focus given drones and aircraft, but terrestrial vehicles are catching up. More sensors, networks and advanced microcontrollers are being deployed to help warfighters better understand situations and respond quickly. Automated systems handle some tasks so humans can focus on their roles, paving the way to autonomous mobility.

The need to configure ground vehicles for specialized tasks is driving a shift to standards and commercial off the shelf (COTS) technologies that pave the way to plug and play modules. COTS has been discussed for years, but its impact is still just beginning to transform many higher volume vehicles.

“We are encouraging use of open standard interfaces for a new system development. This will allow ground vehicle system to use latest sensor technology assuming it complies with the open interface standard,” said Alex Kade, Chief Architect, Ground Vehicle Robotics, at the U.S. Army TARDEC. “In addition, pre-processing much of the sensor information within the sensing system significantly reduces the burden that the central-autonomy and vehicle-management ECUs have to deal with."

Standards make it easier to put a sensor array on a vehicle. But when several cameras, radar and other systems are all streaming in high volumes of data, it can be challenging for controllers to process data in real time. These controllers must also deal with data that comes in different formats and various data rates.

“Different sensor phenomenologies have their own failure modes, and it helps to fuse the sensors for more robust understanding of the environment,” said David Simon, Lead Systems Architect, Autonomy, at Lockheed Martin Missiles and Fire Control. “Higher performing sensors require additional computing. The computing architectures utilizing massively parallel CPUs are a very good fit for processing large amounts of sensor data.”

Several developers use smart sensors that do some processing before data is sent to the controller, the experts say. That reduces the amount of processing power needed in controllers. This is important because powerful multicore processors and the software that drives them are often far more expensive than the simple processors commonly used in sensors. In high volume vehicles, cost can often be a critical factor. That means many designs for these vehicles use networked modules instead of rack mounted boards.

“Trucks are very cost sensitive, so they’re not likely to use boards and backplanes,” said David Jedynak, CTO at Curtiss-Wright Defense Solutions. “If a single-board computer is $8, you still need $2-$10 for a backplane. It’s more cost effective to buy a black box; we make standalone computers with I/O that cost $8. On larger vehicles, boards and backplanes give you a lot of opportunities to repair and replace boards." For trucks, which usually have only one computer, users will want a box approach, Jedynak said.

These cost concerns ripple down to the smallest components. Military requirements have long been more stringent than automotive requisites, but that may change as safety and reliability demands push automotive requirements towards military levels.

“Standardized connectors are always desirable, but we also need to consider the lifecycle requirements of military trucks vs. commercial,” Kade said. “In the past, this has driven military systems to very expensive connection systems. But as we go forward, I believe there may be an opportunity to communize requirements to a high enough degree, that we would consider letting the auto industry drive this market, perhaps through such entities as USCAR or SAE.”

Sensing intelligence

Military developers are also leveraging the advances in automotive sensors, which are following the trend to more functionality for lower prices. Most of the sensing technologies used in safety systems and autonomous cars provide data that’s useful for military users.

“We see that commercial automotive is driving the development of smarter sensors. Lockheed Martin tracks and takes advantage of this if possible,” Simon said. “However, there are some interesting start-up companies that are looking at improving radar and providing raw data. We are eager to see those developments as they mature.”

This choice between distributed intelligence in sensors and powerful centralized modules that process raw data presents an ongoing challenge for development teams. Many design engineers say far more computing power will be needed in the future as the level of automation grows. Many developers predict that artificial intelligence will be needed to analyze all the data coming from sensors.

“I believe architecture will evolve into having powerful centralized computing platforms that will perform as the AI for automated systems, with a precise view of the surrounding environment, vehicles, people, things, etc.,” Kade said. “There will also be a lot of distributed intelligence for sensor pre-processing and reactive ‘safing’ of the autonomous system, providing the appropriate action/reaction with built-in safety protocol and limits.”

Though there’s been a trend to embed processors in sensors, the dramatic increase in computing power is prompting more users to rely on powerful multicore CPUs that process unaltered data from sensors. That can make it easier to synch inputs from multiple sensors while also reducing the possibility that something gets lost when sensors perform processing and send edited files.

Multi-role cores

Advances in microcontrollers impact several facets of military development programs. The rapid increase in computing cores lets engineers dedicate cores to specific tasks. That makes it easier to run virtual machines, so different operating systems can be used. Multicore chips also keep power consumption to manageable levels.

“Processors like Intel’s Xeon have 12 processor cores and power consumption similar to something that’s put in a laptop,” said Mike Southworth, Product Manager at Curtiss-Wright Defense Solutions. “That number of cores opens the possibility of using virtual machines that run on separate cores.”

Power budgets grow in importance as vehicles add more electronic functions. Power budgets can be critical when vehicles are being upgraded with advanced systems. Conventional batteries and wiring systems may have to give way to higher voltages to meet growing demands.

“We have been able to integrate required sensors, by-wire kit, and computer into existing Army trucks without impacting its baseline power demands, but this will be a constant balance between upgrading sensing, actuation, computing power vs. component efficiency improvements,” Kade said. “Eventually, we may need to beef up the electrical system, but as our vehicles become more and more electrified, this may occur anyway as a matter of course.”

Continue reading »