Descripción del proyecto
Un paso artificial por delante para la tecnología de los vehículos autónomos
A raíz de los recientes avances en el campo del aprendizaje artificial, muchos tratan de utilizar esta tecnología en otros ámbitos. Uno de los campos que muestra un gran potencial es el sector de los vehículos autónomos (VA). Estos coches sin conductor podrían proporcionar ventajas importantes para el sector de los transportes, ya que permiten reducir el tráfico, los accidentes, así como los costes y tiempos de desplazamiento. Por desgracia, las tareas actuales de aprendizaje profundo necesarias para que los VA funcionen de forma eficaz y segura consumen demasiada energía y potencia de procesador para los modelos actuales. El proyecto Hailo-8, financiado con fondos europeos, se propone desarrollar una alternativa llamada Hailo-8 que consumiría menos energía, ocuparía menos espacio y engendraría menos costes a la vez que permitiría el avance de las tecnologías para VA.
Objetivo
Autonomous Vehicles (AVs) present a great opportunity for the transport sector to reduce accidents, traffic congestion, time of travel and travel costs. However, for effectiveness, AVs need to process large amounts of data collected by the vehicle sensors at the edge, which requires a very powerful processor capable of computing Deep Learning (DL) tasks. This is currently lacking in the market as evidenced by the inefficiencies in current processors in processing big data at the edge in real time. Most processors for edge computing are currently reliant on CPU and GPU architectures which are challenged by Deep Learning tasks. The processors have low computational capabilities which increases their latencies (processing times). This leads to heat dissipation problems and high power consumption. The processors are also rigged with complexities that raise development costs and the price of the processors. The processors are also not easily scalable, which makes it difficult for miniaturisation.
Hailo-Tech has developed Hailo-8, which is specifically designed to optimise Edge Computing processor capabilities to allow neural network deployment through enhancing processor computational efficiency, resulting in higher capacity within the constraints of an edge device. Hailo-8 meets the industry need of optimised edge data processing by providing a first-class ASIC micro-processor that is based on a completely new micro-architecture that can execute neural network based machine learning algorithms. Hailo-8 will provide AV owners with high computational efficiency (x1,000 compared to alternative solutions), giving an immediate response after data processing. Hailo-8 increases power efficiency by a factor of 100 and has better area and cost efficiency by a factor of 10 compared to other processors. To bring the disruptive device successfully to the market we need to further perform some technical and commercial activities which required an investment of €2.993,750 M.
Ámbito científico
Not validated
Not validated
- engineering and technologymechanical engineeringvehicle engineeringautomotive engineeringautonomous vehicles
- engineering and technologyelectrical engineering, electronic engineering, information engineeringelectronic engineeringcomputer hardwarecomputer processors
- natural sciencescomputer and information sciencesartificial intelligencemachine learningdeep learning
- natural sciencescomputer and information sciencesdata sciencedata processing
- natural sciencescomputer and information sciencesartificial intelligencecomputational intelligence
Programa(s)
Convocatoria de propuestas
Consulte otros proyectos de esta convocatoriaConvocatoria de subcontratación
H2020-SMEInst-2018-2020-2
Régimen de financiación
SME-2 - SME instrument phase 2Coordinador
6789139 TEL-AVIV
Israel
Organización definida por ella misma como pequeña y mediana empresa (pyme) en el momento de la firma del acuerdo de subvención.