SIEMENS validation tools help rapidly develop safer, more reliable vehicles

Introduction

The car of the future will drive the convergence of many technologies. Electrification, sensors, connectivity, cloud computing, big data and AI are closely linked in autonomous vehicles, vehicle-to-everything (V2X) communications, and the functional safety of infotainment electronics and their driver assistance features.

Furthermore, these vehicles are the ultimate system of systems of systems. At the lowest level, we have individual sensors and integrated circuits. They interact within the vehicle’s subsystems, which in turn make up the vehicle itself. But the system doesn’t stop there: the vehicle is just part of an overall vehicular environment that includes other vehicles, pedestrians, infrastructure, and even the cloud.

Figure 1. The vehicle is a system of systems of systems.

This makes the validation of automotive systems a daunting task. The number of scenarios that need to be examined is literally in the millions, and each scenario has different variations. For example, in one scenario, a car may be approaching a pedestrian in a crosswalk. But this may happen at different times of the day, with different weather, different pedestrian clothing, and different ethnicity. In fact, all these factors together make for a validation project that is absolutely impossible to complete with manual physical methods.

At the 2016 Paris Motor Show, Toyota CEO Akio Toyoda reminded us that “14.2 billion miles of testing are needed to validate a vehicle.” In a 2014 article, Autonomous Driving, Roland Berger noted that “design validation is a very important part, if not the most expensive part.” McKinsey’s report, When Will the Robots Hit the Road, warned that while hardware innovation will be fruitful, software will remain a key bottleneck. “Danger can come at any time if anyone fails to fully appreciate the difficulty of automotive design. In this industry, only by being fully prepared can you avoid being caught off guard by some of the hard work.”

The complexity of automotive design is high enough, but safety and security make it even more complex because it is a matter of life and death. Certifications are emerging as standards such as ISO 26262 and the upcoming SOTIF (Safety of Intended Functionality) standard, which aims to define test scenarios, mature. Suppliers cannot afford to play fast and loose when it comes to safety; they must prove it.

In addition to all the above needs, bringing cost-competitive products to market as quickly as possible is another common challenge. This problem urgently requires verification tools that can improve the manageability of this process. To get a new car on the road quickly and efficiently, a combination of realistic scenario modeling , hardware accelerated simulation, and electromechanical verification must be used.

Three car components

For self-driving cars to be effective, their systems must perform three tasks:

• Perception: The vehicle must be able to sense its surroundings. In addition, it must sense many internal conditions for proper operation.

• Computation: The output of these sensors must be evaluated in order to make decisions.

•Execution: These decisions must control some part of the vehicle or some aspect of its operation.

Any comprehensive validation process needs to include all three elements mentioned above. This presents a serious challenge as there is no time to use trial and error to find problems through physical prototyping. And we certainly cannot perform complete safety and security tests in real physical vehicles. The only way to perform a comprehensive validation exercise is to virtualize the entire system, including the environment and the vehicle.

This means we need tools to perform the following tasks:

•Verify real-world environmental conditions and sensor outputs in response to those conditions.

•Verify the circuitry that performs the decision-making calculations given the sensor inputs.

• Make calculated decisions and apply them to virtualized versions of the mechanical systems those decisions control.

Modeling the driving environment

The PreScan tool from the Siemens Tass division performs the first task. It extensively models vehicle infrastructure (such as roads or road sections, bridges and intersections), physical objects (such as trees, buildings and traffic signs), other vehicles and pedestrians, and weather conditions . It also has a comprehensive library of modeled sensors, including cameras, radars, lidars, ultrasonic sensors, infrared sensors, V2X communications, and GPS.

These elements work together to allow for modeling of realistic road conditions, with model variations for time of day, weather, color of vehicle or pedestrian clothing, pedestrian characteristics, and many other methods that can be used to test these scenarios. Together, these virtual scenarios produce the signals that the various vehicle sensors would generate as they react to the scenarios. These signals can then be used to test the integrated circuits responsible for responding to the sensors.

Figure 2. The Siemens Tass group's PreScan tool is capable of modeling a wide range of real-world scenarios.