TDA2X, a SoC optimized for advanced driver assistance systems
TDA2X, a SoC optimized for advanced driver assistance systems
复制标题
TDA2X，一款针对高级驾驶辅助系统进行优化的 SoC
DOI：
--
复制
发表时间：
2014
期刊：
影响因子：
--
通讯作者：
Z. Nikolic
中科院分区：
文献类型：
--
作者：
Jagadeesh Sankaran;Z. Nikolic
TDA2X是德州仪器的一款优化的可扩展片上系统（SoC）解决方案，涵盖ADAS的各种应用领域，如前置摄像头、环绕视图和新兴的传感器融合领域。它通过一组集中的异构处理器来实现这一目标，这些处理器在一个可扩展的架构中与一组丰富的集成外设结合在一起，以低功耗为高级驾驶员辅助系统（ADAS）视觉分析提供最佳的性能组合。各种ADAS应用系统中的计算机视觉算法在处理要求方面具有丰富的变化和多样性沿着需要在具有挑战性的热预算内同时运行它们。具有各种可编程元件的异构体系结构允许系统开发者将算法的各个部分映射到最适合于底层任务的体系结构，从而允许最大化系统性能并减少开发这些复杂系统的开发时间和努力。通过改变这些异构体系结构的内核数量和时钟速度来实现体系结构的可扩展性，允许通过一次软件投资在低、中、高端产品上实现性能和功耗的可扩展性。考虑到ADAS应用的使命关键性，对内核和各种存储器的功能安全性的关注尤为重要。
TDA2X is an optimized scalable system on chip (SoC) solution from Texas Instruments that spans various application areas of ADAS such as front-camera, surround-view and the emerging area of sensor fusion. It accomplishes this through a focused set of heterogeneous processors, brought together in a scalable architecture with a rich set of integrated peripherals, providing an optimal mix of performance in a low power footprint for Advanced Driver Assistance Systems (ADAS) vision analytics. Computer vision algorithms across the various ADAS application systems have a rich variation and diversity in processing requirements along with the need to run them concurrently within challenging thermal budgets. A heterogeneous architecture with various programmable elements allows system developers to map various portions of the algorithms to the architectures that are best suited for the underlying task allowing maximizing system performance and reducing development time and effort in developing these complex systems. Scalability of the architecture by varying the number of cores and clock speeds of these heterogeneous architectures, allows for scalability in performance and power across low, mid and high end products with one software investment. A critical focus on functional safety across the cores and various memories is particularly essential given the mission critical nature of ADAS applications.