Learning efficient millimeter wave radar imaging for autonomous vehicles
The emerging autonomous vehicle ecosystem is expected to grow with an almost 40% CAGR in the next decade hitting €485 billion by 2026 and exceeding €6 trillion in 2050. There is wide industry consensus that improved long-range dep...
ver más
¿Tienes un proyecto y buscas un partner? Gracias a nuestro motor inteligente podemos recomendarte los mejores socios y ponerte en contacto con ellos. Te lo explicamos en este video
Información proyecto DEEP-RADAR
Duración del proyecto: 18 meses
Fecha Inicio: 2019-09-18
Fecha Fin: 2021-03-31
Fecha límite de participación
Sin fecha límite de participación.
Descripción del proyecto
The emerging autonomous vehicle ecosystem is expected to grow with an almost 40% CAGR in the next decade hitting €485 billion by 2026 and exceeding €6 trillion in 2050. There is wide industry consensus that improved long-range depth sensing modalities are imperative for the viability of self-driving cars. State-of-the-art optical technologies are still prohibitively expensive, have insufficient temporal and spatial resolution, do not provide an accurate velocity reading, and are restricted to a shorter range in adverse weather conditions. Millimeter wave multiple-input multiple-output (MIMO) radars are an attractive alternative relying on a phased array of transmitting antennas and digital receivers, containing no moving parts, and able to penetrate adverse weather conditions. The weakness of this technology is the costly requirement for a large number of receiver channels to achieve sufficient spatial resolution. We will apply our novel methodology recently developed for medical imaging to overcome this challenge.
We have demonstrated that learning the entire imaging pipeline in medical ultrasonography, including the shape of the transmitted pulses and the configuration of the receivers allows reducing the number of transmits by a factor of 3, while maintaining image quality comparable to traditional high-frame rate imaging protocols. Despite the different underlying physics, ultrasound and radar imaging share many conceptual similarities and have a similar mathematical description. Here, we intend to develop a proof-of-concept MIMO radar system demonstrating that by using the learned transmit patterns and image reconstruction pipeline, it is possible to halve the number of receive channels without compromising the image resolution and signal-to-noise ratio. Maintaining high resolution images using a smaller number of receiver channels will significantly reduce the cost of this technology and increase the commercial viability of automotive MIMO radars.