Regulatory bodies are beginning to mandate the deployment of ADAS systems, much as they mandated seatbelts and airbags in the 20th century. For example, in an effort to reduce back-over accidents, the U.S. National Highway Transportation Safety Authority (NHTSA) has decreed that, by May 2018, all new vehicles under 10,000 pounds must have a rear-view camera. NHTSA is also assessing the safety benefits of autonomous emergency braking, vehicle-to-vehicle (V2V) communications, and other technologies.

Meanwhile, a number of countries, including the U.S., have implemented government-regulated New Car Assessment Programs (NCAPs), which evaluate the crashworthiness of new vehicles. The Euro NCAP, for example, offers an advanced rewards program for ADAS systems that have been scientifically proven to provide a safety benefit to consumers. Case in point: the “Collision Prevention Assist” system in the 2014 Mercedes-Benz V-Class van, which received a five-star safety rating from the Euro NCAP. The system combines radar-based forward-collision warnings with adaptive brake assist, which applies the optimum braking force to avoid or mitigate a forward collision while also ensuring that vehicles behind have a chance to stop safely.

Active safety

As the automotive industry continues on the path to autonomous driving, systems that offer “passive” assistance by alerting drivers to potential danger, such as a motorcycle in the vehicle’s blind spot, have evolved into “active” safety systems (Figure 1). These typically discrete systems can control steering, braking, or engine throttle, using input from radar or camera sensors. Next-generation active systems now in development will be much more effective at preventing crashes. They will integrate multiple heterogeneous networked sensors and use sophisticated decision-making algorithms to interpret the state of the vehicle and its surroundings more accurately. The systems will use new, purpose-built ADAS SoCs that provide support for the multiple sensor technologies. The wide availability of these deep-submicron SoCs, built by companies such as Freescale, Intel, NVIDIA, Renesas, and Texas Instruments, will spur system designers to replace discrete, single-function ADAS systems implemented on low-end microcontrollers with smarter, consolidated multi-sensor systems that can handle high data throughput in real time.

 

 

The new SoCs can manage a large amount of data from multiple streams, including vision, infrared, LiDAR, ultrasonic, and radar (long, medium, and short-range), allowing the ADAS system to achieve highly accurate detection and recognition of pedestrians, vehicles, and other objects (Figure 2). These SoCs often incorporate multiple CPU cores, digital signal processors, general-purpose graphics processing units (GPGPUs), or vision acceleration engines, along with multiple camera inputs and display outputs. The SoCs will offer automakers increased flexibility and control over how their ADAS systems are designed, enabling commercially attractive and differentiated solutions. Rather than adopt a shrink-wrapped hardware/software system as many do today, automakers will have the freedom to leverage a full ecosystem of ADAS suppliers, with best-of-breed technologies making their way to the forefront – much the same way that infotainment systems evolved.

 

 

Sensor fusion

ADAS systems must function under a variety of weather and lighting conditions. A vision-based system, for example, should have the intelligence to detect a poor visibility condition, such as snow, heavy fog, or sunlight shining directly into the lens. The system could then disable itself and warn the driver that it is non-operational, or, better yet, prioritize complementary sensor data, such as radar for forward-collision detection, that isn’t affected by inclement weather. Another example is an ultrasonic parking sensor that becomes prone to false positives when encrusted with mud. In that case, the system could disregard ultrasonic sensor data and use short-range radar or camera data instead. By combining the results of different sensors or different sensor technologies through sensor fusion, system designers can create a more effective solution than by using a single technology in isolation.

As systems become more integrated and present more data to the driver, they can potentially cause driver information overload. This overload could in turn result in a high cognitive workload, reducing situational awareness and mitigating the effectiveness of ADAS. System designers must therefore devise easy-to-use systems that make use of the most appropriate modalities (visual, manual, tactile, aural, haptic) for the task at hand. To ensure an optimal user experience, designers must also establish a clear specification of the driver-vehicle interface to ensure the correct balance of user and system requirements.

Testing and validating ADAS systems is, arguably, one of the greatest challenges automakers face. Prior to deploying a commercial ADAS system, the development team must amass hundreds, if not thousands, of scenarios for testing in a regression test database, in an effort to test all scenarios. The ultimate goal is to achieve 100 percent accuracy and zero false positives under all possible conditions, regardless of traffic, weather, or number of obstacles or pedestrians in the scene. But how can the team be sure that the test database comprises all test cases? The reality is they cannot – which is why suppliers spend years testing and validating systems, and performing extensive field trials in various geographies prior to commercial deployment.

A matter of compliance

Compliance with multiple safety-related standards has become a minimum requirement for ADAS systems. It allows an automaker and its suppliers to demonstrate that they use consistent, auditable processes in their product development and manufacturing, and in their risk assessment of hazardous operational scenarios. For example, compliance with ISO 26262, an automotive adaptation of the IEC 61508 functional safety standard, demonstrates that a system is designed, implemented, and maintained to meet its automotive safety integrity level (ASIL), which can range from level A (representing the lowest degree of hazard) to D (representing the highest). The system manufacturer determines the ASIL according to the severity, probability, and controllability of the risks involved in the system.

Unlike IEC 61508, ISO 26262 doesn’t declare precise values for acceptable probabilities of failure. Rather, it assesses risks in a qualitative fashion and defines safety measures to avoid, control, or mitigate the effects of systematic or random failures of the overall system. The standard covers all parts of the automotive supply chain, including the SoCs, operating systems, middleware components, and algorithms developed by Tier 2 suppliers; the hardware and software developed by Tier 1 suppliers; and the final systems built by OEMs.

Safety certification of ADAS systems can be a long and arduous process. Using pre-certified components can make system-level certification much easier and contribute to a greater level of safety. The QNX OS for Automotive Safety, for example, has been pre-certified for use in ADAS systems that comply with ISO 26262, up to ASIL D. This certification makes the OS suitable for a wide variety of ADAS systems, from PRNDL displays to pedestrian avoidance systems.

Besides ISO 26262, ADAS systems may need to comply with additional criteria dictated by the tier 1 supplier or automaker. On the software side, these criteria may include AUTOSAR (a standardized automotive software architecture for electronic control units), OpenCL (a software framework that simplifies parallel computing tasks), or MISRA (vehicle-based software development guidelines for the C and C++ languages). On the hardware side, ADAS systems need to pass AEC Q100 qualification, which involves reliability testing of automotive-grade integrated circuits (ICs) at various temperature grades. ICs must function reliably over temperature ranges that span from -40 °C to +150 °C, depending on the system.

Stretching the model

While legislation, industry standards, and technology advancements are all helping the ADAS market reach economies of scale, each of these factors must continue to evolve to address the demands of a changing industry. For example, the emergence of SoCs with multiple ARM cores, GPUs, or vision acceleration cores may stretch the traditional AUTOSAR software model beyond its current capabilities. Ensuring optimal inter-core communication and shared-resource utilization in a multicore environment is a complex problem to solve. Add to that the requirements imposed by ISO 26262, and a complete overhaul of AUTOSAR or perhaps even a new generation of OS with established inter-process communication for processes running on different cores may be required.

Before ADAS systems can become as commonplace as the steering wheel, the industry must also address other challenges, including a lack of interoperability specifications for radar, laser, and video data in the car network. For audio/video data alone, automakers use multiple communication standards, including Ethernet AVB, LVDS, and MOST. Consequently, ADAS systems must support a multitude of interfaces to ensure widespread adoption. They may also need additional interfaces for V2V and vehicle-to-infrastructure (V2I) data. The industry needs working models that will enable complementary, redundant sensors to work in concert and thereby increase the efficacy of ADAS solutions.

The road to adoption

ADAS systems are commercially available today and next-generation systems promise great technological advancements. Consumer demand is high and the road to widespread adoption is being paved. In fact, Visiongain forecasts that the global ADAS market will experience double-digit growth between 2014 and 2024, from a baseline estimate of $18.2 billion[1]. Some hurdles to bringing next-generation ADAS systems to mainstream vehicles remain, but given the remarkable strides made over the last few years and the burgeoning ADAS ecosystem, the industry can and will overcome these challenges.

Reference:

[1] Visiongain: Automotive Advanced Driver Assistance Systems (ADAS) Market 2014-2024.

Tina Jeffrey is an automotive product manager at QNX Software Systems, where she supports infotainment, ADAS, and acoustics products. Tina received her BS in electrical engineering from the University of Ottawa.

QNX Software Systems

www.qnx.com

Twitter: @QNX_News

LinkedIn: opsy.st/QNXLinkedIn

Facebook: www.facebook.com/QNXSoftwareSystems

YouTube: www.youtube.com/user/QNXcam