AImotive will showcase its concepts for the future of automated driving on embedded at Embedded World 2019 in Nüremberg, Germany. Key modules of the recently announced next-generation modular self-driving software stack, aiDrive, will be demonstrated running on the Nvidia Drive PX2 platform.
AImotive’s new vehicle system simulator aiSim, launched at CES 2019, will also be on display featuring a new purpose-built physics-accurate simulation and rendering engine. A silicon implementation of AImotive’s aiWare hardware IP will also be shown, developed by Verisilicon using GlobalFoundries’ 22FDX process, demonstrating the potential power and performance benefits of highly optimized NN hardware acceleration.
To emphasize its drive for maximizing collaboration in the automated driving industry, AImotive has designed its toolbox of technologies for the development of self-driving technologies in a modular and highly scalable way. This allows OEMs, Tier1s and robotaxi providers to mix and match AImotive’s technologies with their existing solutions to create more advanced, cost-competitive and safer automated driving systems, due to which the company’s partners and customers are able to utilize the modular architecture of aiDrive, and the aiSim verification platform, to create productizable automated driving systems faster.
Beyond aiDrive this is showcased in the aiSim2 simulator. The new simulator can drastically accelerate the development of autonomous technologies while overcoming the limitations of game engine-based simulators through the ability to ensure deterministic physically-based rendering on any hardware setup. Due to its modular architecture selected modules of aiSim can also easily be integrated into other existing simulation solutions.
The company’s unique neural network accelerator hardware IP, aiWare, will also be on show at Embedded World. Created through a partnership between AImotive, VeriSilicon and Global Foundries, the chip accelerates any CNN, including AImotive’s own proprietary networks, while consuming 10x-20x less power than a state-of-the-art GPU executing the same task at the same speed. The highly scalable aiWare architecture can be licensed for a wide range of embedded use cases, including smart sensor, sensor fusion engines, or central compute clusters for automated driving platforms.