Smarter Sensor Design for Self-Driving Cars

Smarter Cars Rely on Smarter Sensors

A variety of sensors are used in today’s cars. Radar, LiDAR, image, pressure, and ultrasonic are just a handful of examples. Sensors collect data from in and around the vehicle. The data is then sent to a controller consisting of a processor and components which make up an SoC, which runs algorithms that turn this data into actionable insights that drive the vehicle’s actions. For example, image sensors support advance driver assistance systems (ADAS), which provide capabilities such as lane departure warnings, blind spot detection, and automatic braking. Similarly, ultrasonic and radar sensors enable alerts for object proximity, so you do not hit something.

To be effective, sensors need to:

• Have a small footprint
• Consume minimal power
• Deliver defined performance capabilities
• Operate in harsh environments
• Comply with functional safety requirements
• Function for an extended period
• Meet stringent cost considerations

While many of today’s electronic products also call for small, power-efficient, and budget-friendly components, their similarities with their automotive counterparts end there. The environments in which cars operate bring a new dimension to the mix that must account for factors like temperature extremes, humidity, and electrical noise. What’s more, the average age of vehicles on the road is approaching 12 years, according to research from IHS Markit. Based on this and safety issues, automotive manufacturers expect that the semiconductors they use will have a zero parts per billion failure rate for 15 years.  Component-related challenges aside, another important consideration is the design cycle itself. Depending on its complexity, a vehicle design process can take anywhere from a few to several years. However, automotive designers are seeking ways to reduce this timeframe so that they can get their vehicles to market faster.

Holistic Approach Needed to Accelerate the Automotive Design Process

A holistic approach to automotive component design has proven to be beneficial to addressing the key design challenges. This can be achieved by working with an electronic design automation (EDA) vendor that can provide guidance and a robust flow to support every step in the design process. As a starting point, it’s helpful to address the functional safety criteria for each component, as that will determine the paths that should be taken. ISO 26262 is the automotive standard that mandates the development process, from specification through production release and end-of-life, that automotive OEMs and suppliers must follow and document for their devices to be considered functionally safe (i.e., the components will perform as intended). Automotive Safety Integrity Level (ASIL) provides the risk classification system outlined by ISO 26262. For example, a radar sensor that’s part of an ADAS will need to meet the requirements of the most stringent ASIL: ASIL D. By contrast, a light sensor for an interior lighting application is not as crucial to overall vehicle safety, so ASIL B may be sufficient.

The Synopsys team of automotive experts brings together EDA tools, automotive-grade IP, and automotive solutions to help designers create smarter, safer, more secure chips and software faster. Using our safety architecture as well as ASIL planning tools, our experts can help plan for chips to achieve their target ASIL. Then, at the silicon design phase, we have a variety of synthesis, simulation, and physical implementation tools that address “what if?” scenarios. When the design is ready for verification and validation, we have tools to help assess whether the chips have met the targeted ASIL levels. And, to reduce design time and effort, we also have certified ASIL-ready IP assessed for ISO 26262 random hardware faults that support safety-critical applications. While sensors represent a simpler design than SoCs, they can be designed using a similar flow.

There’s also a design approach from Synopsys that can help shorten the development cycle and, ultimately, the vehicle’s time to market, while enabling designers to meet consumer demands and safety standards. We call this approach Triple Shift Left. By turning a traditionally linear automotive development process into a parallel one, Triple Shift Left empowers designers to find problems up to 18 months earlier (before hardware becomes available). Through virtual models, it also provides OEMs visibility into the earliest parts of the design. Additionally, the approach allows designers to build security and quality into their software during development and testing.

Here are the three key steps of Triple Shift Left:

• Shift Left I: Use automotive-grade IP to implement dedicated functions of silicon, like safety islands and processors, with embedded vision, neural network, and sensor fusion capabilities
  
• Shift Left II: Use simulation and shared models to develop software early on a virtual platform, which allows parallel software and hardware development
• Shift Left III: Increase coverage and accelerate test cycles for applications, power electronics, and wire harness simulations; use automated regression to deliver safer, more reliable software.

As with so many other things in our lives, from everyday gadgets to mission-critical systems, our cars are getting smarter, more connected, safer, and more secure. Designing the sensors and other components that fuel this intelligence and connectivity calls for a tool flow, automotive-grade IP, software testing, and approach that is robust, efficient, and safety-focused. The ultimate goal is to help designers more quickly build smarter, more secure chips and software for safety-critical automotive applications.