Improving safety standards for autonomous vehicles

In the near future, fully autonomous vehicles are expected to significantly improve automotive safety and transportation efficiency. However, to achieve this goal, automotive OEMs must go beyond the current level of automotive autonomy. To achieve this leap, they need to overcome a series of unique challenges related to testing automotive radar sensors in advanced driver assistance systems (ADAS) and autonomous driving systems, and develop new methods to train algorithms that traditional solutions cannot handle.

The Society of Automotive Engineers (formerly the American Society of Automotive Engineers) divides autonomous driving into six levels, with L0 representing full driver control and L5 representing fully autonomous driving.

Today’s most advanced autonomous vehicle systems are only at Level 3, which means the vehicle can make decisions such as accelerating or braking without human intervention. Going from Level 3 to Level 5 requires breakthroughs in many areas, including bridging the gap between software simulation and road testing, and training ADAS and autonomous driving algorithms under real conditions.

Keysight Technologies' newly released Radar Scenario Emulator (RSE) goes a long way toward bridging these gaps.

Software simulation plays an important role in the development of autonomous vehicles. Software simulation environments help manufacturers validate the capabilities of ADAS and autonomous driving systems. However, simulation cannot fully replicate real-world driving conditions or sensor malfunctions, which are inevitable issues for fully autonomous vehicles.

Before bringing ADAS and autonomous driving systems to market, OEMs verify their reliability through road testing. While road testing is and will continue to be a critical and integral part of the development process, it is time-consuming, expensive, and difficult to repeat in terms of controlled environmental conditions. If relying solely on road testing, it would take decades to develop a reliable vehicle that can safely operate on urban and rural roads at all times. Training algorithms is essential to complete development within a relatively realistic timeline.

Validating radar-based autonomous driving algorithms is a critical task. Sensors collect information about road and traffic conditions and provide this information to processors and algorithms so that they can make decisions about how the vehicle should respond in a given situation. Without proper training, autonomous vehicles can make decisions that could put the safety of their occupants or pedestrians at risk.

Becoming a good driver often requires time and experience. Autonomous driving systems also need long-term training to improve their ability to cope with real-world driving conditions. Achieving Level 5 autonomous driving requires complex systems that are more capable than the best human drivers.

There are also risks in prematurely testing unproven ADAS and autonomous driving systems on the road. OEMs need to be able to simulate real-world scenarios to validate the actual sensors, ECU code, AI, etc.

Current lab-based simulation solutions do not provide driving scenarios that are close to real-world conditions. They have a limited field of view and cannot resolve objects closer than 4 meters. Some systems use multiple radar target simulators (RTS), each of which presents multiple point targets to the radar sensor and simulates horizontal and vertical positions by mechanically moving the antenna. This mechanical automation slows down the overall test speed. Other solutions use only a few target simulators to form an antenna wall, and objects can appear anywhere in the scene, but they cannot appear at the same time. In a static or quasi-static environment, this approach can test a few targets moving laterally, but is limited by the speed of the robotic arm.

Current simulators can only simulate up to 32 objects, including vehicles, infrastructure, pedestrians, obstacles, and other objects. This is far less than the number of objects a vehicle on the road might encounter at any time. When testing radar sensors, a limited number of objects cannot reflect complete driving scenarios and cannot reproduce the complex situations in the real environment.

To advance autonomous driving technology to L4 and L5 , automotive OEMs need solutions that can render more objects faster and at closer distances. To meet this need, Keysight has developed a dedicated scalable simulation screen. This screen consists of hundreds of micro-target radar simulators and can simulate up to 512 objects as close as 1.5 meters. As a result, the simulation screen can present a deterministic real environment for testing complex scenarios in the lab that could only be tested on the road before.

Keysight is very proud to have made breakthroughs in these technologies, resulting in the launch of the Radar Scenario Simulator product, which is a key component of the Automated Driving Simulation (ADE) platform developed by Keysight. We believe that this technology can reduce road traffic casualties, ease traffic congestion, and enable safer and more efficient road traffic, thus shortening the distance to Level 5 autonomous driving.