| Reference : Vehicle-to-Infrastructure Communication for Real-Time Object Detection in Autonomous ... |
| Scientific congresses, symposiums and conference proceedings : Paper published in a book | |||
| Engineering, computing & technology : Computer science | |||
| http://hdl.handle.net/10993/53300 | |||
| Vehicle-to-Infrastructure Communication for Real-Time Object Detection in Autonomous Driving | |
| English | |
Hawlader, Faisal  [University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT) > PI Engel >] | |
| Robinet, François [University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT) > SEDAN >] | |
| Frank, Raphaël [University of Luxembourg > Interdisciplinary Centre for Security, Reliability and Trust (SNT) > SEDAN >] | |
| Jan-2023 | |
| 18th Wireless On-demand Network systems and Services Conference (WONS-23) | |
| Yes | |
| Yes | |
| International | |
| 18th Wireless On-demand Network systems and Services Conference (WONS-23) | |
| 30-01-2023 to 01-02-2023 | |
| [en] 5G ; Cloud/Edge Computing ; Perception ; C-V2X ; Autonomous Driving | |
| [en] Environmental perception is a key element of autonomous driving because the information receive from the perception module influences core driving decisions. An outstanding challenge in real-time perception for autonomous driving lies in finding the best trade-off between detection quality and
latency. Major constraints on both computation and power have to be taken into account for real-time perception in autonomous vehicles. Larger object detection models tend to produce the best results, but are also slower at runtime. Since the most accurate detectors cannot run in real-time locally, we investigate the possibility of offloading computation to edge and cloud platforms, which are less resource-constrained. We create a synthetic dataset to train an object detection model and evaluate different offloading strategies. Using real hardware and network simulations, we compare different trade-offs between prediction quality and end-to-end delay. Since sending raw frames over the network implies additional transmission delays, we also explore the use of JPEG compression at varying qualities and measure its impact on prediction metrics. We show that models with adequate compression can be run in real-time on the cloud while outperforming local detection performance. | |
| Researchers ; Professionals ; Students ; General public | |
| http://hdl.handle.net/10993/53300 |
| File(s) associated to this reference | ||||||||||||||
|
Fulltext file(s):
| ||||||||||||||
All documents in ORBilu are protected by a user license.
