Saved in:
| Main Authors: | , |
|---|---|
| Format: | Recurso digital |
| Language: | |
| Published: |
Zenodo
2026
|
| Online Access: | https://doi.org/10.5281/zenodo.18289635 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866902189581008896 |
|---|---|
| author | Reddy Chandrashekar, Nithik K, Vinay |
| author_facet | Reddy Chandrashekar, Nithik K, Vinay |
| contents | <p>The evolution of autonomous vehicles has placed unprecedented demands on embedded computing systems, particularly in visual perception. Cameras and other vision sensors continuously generate vast amounts of data that must be processed in real time to ensure safe navigation and accurate environmental understanding. Conventional computing platforms, such as CPUs and GPUs, often struggle to meet these requirements due to their high-power consumption and limited determinism under strict automotive constraints. Vision Processing Units (VPUs) implemented through advanced Very Large-Scale Integration (VLSI) design techniques have emerged as a powerful alternative, offering specialized architectures optimized for low-latency, high-throughput, and energy-efficient visual computation.</p> <p>This paper presents a comprehensive study on the VLSI implementation of VPUs tailored for autonomous vehicle applications. It examines architectural principles, hardware–software co-design strategies, and optimization techniques aimed at improving performance while adhering to functional safety and reliability standards such as ISO 26262. The proposed design incorporates parallel processing elements, reconfigurable logic blocks, and an optimized on-chip memory hierarchy to efficiently execute vision algorithms, including convolutional neural networks and feature extraction tasks. Emphasis is placed on balancing flexibility and specialization so that the hardware can adapt to evolving perception algorithms without sacrificing efficiency. Simulation and synthesis analyses indicate that the proposed architecture achieves substantial gains in processing speed and energy efficiency compared to conventional systems. Overall, this research demonstrates that custom VLSI-based VPUs can serve as a cornerstone technology for next-generation autonomous vehicles, enabling intelligent, power-aware, and functionally safe perception systems.</p> |
| format | Recurso digital |
| id | zenodo_https___doi_org_10_5281_zenodo_18289635 |
| institution | Zenodo |
| language | |
| publishDate | 2026 |
| publisher | Zenodo |
| record_format | zenodo |
| spellingShingle | VLSI Implementation of Vision Processing Units for Autonomous Cars Reddy Chandrashekar, Nithik K, Vinay <p>The evolution of autonomous vehicles has placed unprecedented demands on embedded computing systems, particularly in visual perception. Cameras and other vision sensors continuously generate vast amounts of data that must be processed in real time to ensure safe navigation and accurate environmental understanding. Conventional computing platforms, such as CPUs and GPUs, often struggle to meet these requirements due to their high-power consumption and limited determinism under strict automotive constraints. Vision Processing Units (VPUs) implemented through advanced Very Large-Scale Integration (VLSI) design techniques have emerged as a powerful alternative, offering specialized architectures optimized for low-latency, high-throughput, and energy-efficient visual computation.</p> <p>This paper presents a comprehensive study on the VLSI implementation of VPUs tailored for autonomous vehicle applications. It examines architectural principles, hardware–software co-design strategies, and optimization techniques aimed at improving performance while adhering to functional safety and reliability standards such as ISO 26262. The proposed design incorporates parallel processing elements, reconfigurable logic blocks, and an optimized on-chip memory hierarchy to efficiently execute vision algorithms, including convolutional neural networks and feature extraction tasks. Emphasis is placed on balancing flexibility and specialization so that the hardware can adapt to evolving perception algorithms without sacrificing efficiency. Simulation and synthesis analyses indicate that the proposed architecture achieves substantial gains in processing speed and energy efficiency compared to conventional systems. Overall, this research demonstrates that custom VLSI-based VPUs can serve as a cornerstone technology for next-generation autonomous vehicles, enabling intelligent, power-aware, and functionally safe perception systems.</p> |
| title | VLSI Implementation of Vision Processing Units for Autonomous Cars |
| url | https://doi.org/10.5281/zenodo.18289635 |