Super-parallel optimisation improves simulation performance by 10 times

OTSL Inc, a developer and distributor of short-range wireless and embedded systems, has announced a new 3D real-time millimetre wave radar simulator for autonomous driving (*1), AMMWR2 (Advanced Millimetre Wave Radar Simulator 2). Since the 2017 launch of AMMWR, the world's first sensor simulator software for autonomous driving that enables real-time dynamic simulation, OTSL has continued to invest heavily in development. With the announcement of AMMWR2, OTSL has manifested the leaps in functionality and performance achieved during this development. This product is expected to be marketed globally by the end of this year to automotive manufacturers, system suppliers who develop, design and manufacture vehicle sensors, and semiconductor manufacturers who develop sensor devices.
"With global advances in the application of autonomous driving, expectations are growing for advanced simulation technology capable of virtually reproducing all types of driving conditions and verifying and validating safety and accuracy. However, the current autonomous driving vehicle only has Level 2 or Level 3 capabilities according to SAE International Standards," said Shoji Hatano, CEO, OTSL and OTSL Germany GmbH. "To achieve fully autonomous driving through system monitoring, Level 5 of the standards, it is essential to include simulation of possible accidents caused by electronic device defects and sensor failures. AMMWR2, announced on 17 June, is the only millimetre-wave radar sensor simulator that spans from the electronic device level, including semiconductors and sensors, to the autonomous driving (AD) and advanced driver assistance systems (ADAS) simulation domain."
AMMWR2 uses the ray-tracing method (*2), which is used in the field of optical simulation and unique reflection and diffraction models. It uses a special three-dimensional computer-generated Unreal Engine 4 (EPIC Games, Inc.) map that implements reflection and diffraction models of roads, street trees, street lights, traffic signals, signs and other objects to achieve real-time simulation of dynamic objects. The development of original measurement methods and the creation of reflection models from the data obtained through measurement enables an accurate simulation of the reflection of radar waves, even taking into account the shapes and materials of vehicles, pedestrians, traffic signals, signs and other objects. By strengthening the partnership with COSIDE (R) (*3), a SystemC AMS design and simulation tool from COSEDA Technologies GmbH for semiconductors that supports design from the circuit design to the system design level, an environment can be created in which simulation can be performed entirely in software without the need for any hardware, such as sensors or electronic control units (ECUs), or even the vehicle itself.
AMMWR2 provides the following major update. These performance and functional improvements will enable shorter development and testing lead times for manufacturers working on autonomous driving technologies.
Acceleration of the simulation engine through internal code optimisation and super parallel processing techniques. This has improved the performance of ray-traced simulations by a factor of 10 or more.
By strengthening the partnership with COSEDA Technologies GmbH, which develops the SystemC AMS design and simulation tool COSIDE (R), various radars can be used as custom models, in addition to the unique radar models based on automotive radars that Tier-1 supplies with OEM, including Bosch's FR5CPEC and Continental's ARS540.
AMMWR2 now supports integration with the open-source CARLA simulator, which is widely used for developing and testing algorithms for autonomous driving.
www.otsl.jp

Tweet