Jiajun He
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Jiajun He has a total of 6 co-authors. Here are the top co-authors by number of shared publications:
- Changqing Xu
Guangzhou Institute of Technology, Xidian University, Guangzhou 510555, China
2 shared publications - Xinfang Liao
Guangzhou Institute of Technology, Xidian University, Guangzhou 510555, China
2 shared publications - Yi Liu
Shenzhen Institute of Technology, Xidian University, Shenzhen 518000, China
2 shared publications - Yintang Yang
Xidian University
2 shared publications - Weijie Lu
Guangzhou Institute of Technology, Xidian University, Guangzhou 510555, China
1 shared publication - Yuanhao Zhang
Guangzhou Institute of Technology, Xidian University, Guangzhou 510555, China
1 shared publication
Latest Publications
- Open Access
- 29 January 2026
- Published by MDPI AG in Micromachines
- Vol. 17 (2) , 185
HCCA-SAFE: A Hybrid Cascaded Control Architecture for FPGA-Based Fault Injection in Safety-Critical Automotive SoCs
Automotive System-on-Chips (SoCs) must meet stringent functional safety standards, such as ISO 26262 and IEC 61508, to ensure reliable operation under hardware faults. FPGA-based fault injection has emerged as a practical and cost-effective technique for functional safety verification. However, instrumentation-based methods face scalability challenges when applied to the high fault densities typical of automotive SoCs. To address these challenges, we propose a hybrid cascaded fault-injection controller architecture (HCCA-SAFE) that simultaneously reduces high-fanout global nets and eliminates long serial propagation paths. The architecture constrains enable-signal cluster width and distributes control across cascaded stages, improving timing results and routability under limited FPGA resources. The proposed architecture is evaluated on multiple open-source RISC-V processor cores. On openE902, HCCA-SAFE reduces net delay from 27.276 ns to 22.535 ns and achieves 32.2% and 63.8% lower net delay compared with the representative centralized and shift-chain approaches, respectively. On openE906, the proposed HCCA-SAFE limits the net delay to 12.959 ns and reduces the maximum control-signal fanout to 1763, respectively, compared with 25.825 ns and 40.442 ns in the conventional method. On openC906, the proposed design lowers the maximum control-signal fanout from 7725 to 570 and reduces the net delay to 7.506 ns. Furthermore, HCCA-SAFE produces results fully consistent with software-based RTL simulation, while delivering substantial performance gains. Speed-up factors of , , and are achieved on openE902, openE906, and openC906, respectively, with efficiency improvements scaling with processor complexity These results confirm that HCCA-SAFE delivers scalable, timing-robust fault-injection control suitable for large automotive SoCs.- 29 January 2026
- Published by MDPI AG in Micromachines
- Vol. 17 (2) , 185
Open AccessHCCA-SAFE: A Hybrid Cascaded Control Architecture for FPGA-Based Fault Injection in Safety-Critical Automotive SoCs
Automotive System-on-Chips (SoCs) must meet stringent functional safety standards, such as ISO 26262 and IEC 61508, to ensure reliable operation under hardware faults. FPGA-based fault injection has emerged as a practical and cost-effective technique for functional safety verification. However, instrumentation-based methods face scalability challenges when applied to the high fault densities typical of automotive SoCs. To address these challenges, we propose a hybrid cascaded fault-injection controller architecture (HCCA-SAFE) that simultaneously reduces high-fanout global nets and eliminates long serial propagation paths. The architecture constrains enable-signal cluster width and distributes control across cascaded stages, improving timing results and routability under limited FPGA resources. The proposed architecture is evaluated on multiple open-source RISC-V processor cores. On openE902, HCCA-SAFE reduces net delay from 27.276 ns to 22.535 ns and achieves 32.2% and 63.8% lower net delay compared with the representative centralized and shift-chain approaches, respectively. On openE906, the proposed HCCA-SAFE limits the net delay to 12.959 ns and reduces the maximum control-signal fanout to 1763, respectively, compared with 25.825 ns and 40.442 ns in the conventional method. On openC906, the proposed design lowers the maximum control-signal fanout from 7725 to 570 and reduces the net delay to 7.506 ns. Furthermore, HCCA-SAFE produces results fully consistent with software-based RTL simulation, while delivering substantial performance gains. Speed-up factors of , , and are achieved on openE902, openE906, and openC906, respectively, with efficiency improvements scaling with processor complexity These results confirm that HCCA-SAFE delivers scalable, timing-robust fault-injection control suitable for large automotive SoCs. - Open Access
- 13 June 2025
- Published by MDPI AG in Electronics
- Vol. 14 (12) , 2418
Fine-Grained Fault Sensitivity Analysis of Vision Transformers Under Soft Errors
Over the past decade, deep neural networks (DNNs) have revolutionized the fields of computer vision (CV) and natural language processing (NLP), achieving unprecedented performance across a variety of tasks. The Vision Transformer (ViT) has emerged as a powerful alternative to convolutional neural networks (CNNs), leveraging self-attention mechanisms to capture long-range dependencies and global context. Owing to their flexible architecture and scalability, ViTs have been widely adopted in safety-critical applications such as autonomous driving, where system reliability is paramount. However, ViTs’ reliability issues induced by soft errors in large-scale digital integrated circuits have generally been overlooked. In this paper, we present a fine-grained fault sensitivity analysis of ViT variants under bit-flip fault injections, focusing on different ViT models, transformer encoder layers, weight matrix types, and attention-head dimensions. Experimental results demonstrate that the first transformer encoder layer is susceptible to soft errors due to its essential role in local and global feature extraction. Moreover, in the middle and later layers, the Multi-Layer Perceptron (MLP) sub-blocks dominate the computational workload and significantly influence representation learning, making them critical points of vulnerability. These insights highlight key reliability bottlenecks in ViT architectures when deployed in error-prone environments.
- 13 June 2025
- Published by MDPI AG in Electronics
- Vol. 14 (12) , 2418
Open AccessFine-Grained Fault Sensitivity Analysis of Vision Transformers Under Soft Errors
Over the past decade, deep neural networks (DNNs) have revolutionized the fields of computer vision (CV) and natural language processing (NLP), achieving unprecedented performance across a variety of tasks. The Vision Transformer (ViT) has emerged as a powerful alternative to convolutional neural networks (CNNs), leveraging self-attention mechanisms to capture long-range dependencies and global context. Owing to their flexible architecture and scalability, ViTs have been widely adopted in safety-critical applications such as autonomous driving, where system reliability is paramount. However, ViTs’ reliability issues induced by soft errors in large-scale digital integrated circuits have generally been overlooked. In this paper, we present a fine-grained fault sensitivity analysis of ViT variants under bit-flip fault injections, focusing on different ViT models, transformer encoder layers, weight matrix types, and attention-head dimensions. Experimental results demonstrate that the first transformer encoder layer is susceptible to soft errors due to its essential role in local and global feature extraction. Moreover, in the middle and later layers, the Multi-Layer Perceptron (MLP) sub-blocks dominate the computational workload and significantly influence representation learning, making them critical points of vulnerability. These insights highlight key reliability bottlenecks in ViT architectures when deployed in error-prone environments.