论文信息 - 7.2 A 20.5TOPS and 217.3GOPS/mm2 Multicore SoC with DNN Accelerator and Image Signal Processor Complying with ISO26262 for Automotive Applications

7.2 A 20.5TOPS and 217.3GOPS/mm2 Multicore SoC with DNN Accelerator and Image Signal Processor Complying with ISO26262 for Automotive Applications

Vehicles with features for an advanced driver-assistance system (ADAS) are already available on the market. Technologies for image recognition are essential elements for ADAS applications, and several SoCs performing image recognition for ADAS have been developed [1–3]. As vehicles that support ADAS become more complicated, these technologies need improvement. Recently, deep neural networks (DNNs) have achieved higher recognition accuracy than traditional feature-based matching algorithms, and can be applied to applications such as road detection. Consequently, several image recognition SoCs with DNN processors [4–5] have been developed. Functional safety is important for automotive applications, so several SoCs, such as [2], comply with ISO26262. The requirements for SoCs that support ADAS features are: high performance for detecting various objects, low power use to keep execution stable in the rapidly changing environment of moving vehicles, and high safety to avoid serious accidents.

[1] Hoi-Jun Yoo,et al. A 502-GOPS and 0.984-mW Dual-Mode Intelligent ADAS SoC With Real-Time Semiglobal Matching and Intention Prediction for Smart Automotive Black Box System , 2017, IEEE Journal of Solid-State Circuits.

[2] Nitin Chawla,et al. 14.1 A 2.9TOPS/W deep convolutional neural network SoC in FD-SOI 28nm for intelligent embedded systems , 2017, 2017 IEEE International Solid-State Circuits Conference (ISSCC).

[3] Yasuhisa Shimazaki,et al. 4.5 A 16nm FinFET heterogeneous nona-core SoC complying with ISO26262 ASIL-B: Achieving 10−7 random hardware failures per hour reliability , 2016, 2016 IEEE International Solid-State Circuits Conference (ISSCC).

[4] Desoli Mr Giuseppe,et al. 14.1 A 2.9TOPS/W deep convolutional neural network SoC in FD-SOI 28nm for intelligent embedded systems , 2017 .

[5] Yutaka Yamada,et al. 18.2 A 1.9TOPS and 564GOPS/W heterogeneous multicore SoC with color-based object classification accelerator for image-recognition applications , 2015, 2015 IEEE International Solid-State Circuits Conference - (ISSCC) Digest of Technical Papers.

[6] Hoi-Jun Yoo,et al. UNPU: A 50.6TOPS/W unified deep neural network accelerator with 1b-to-16b fully-variable weight bit-precision , 2018, 2018 IEEE International Solid - State Circuits Conference - (ISSCC).