Private 5G Networks for Real-Time Object Detection in Autonomous Systems
A controlled benchmark of private 5G and Wi-Fi for video streaming and object detection, followed by a separate autonomous forklift demonstration.
The Question
Autonomous systems in industrial settings need real-time video and object detection. They currently rely on Wi-Fi, but private 5G promises dedicated bandwidth, lower interference, and industrial-grade reliability. The question: does 5G actually deliver for real-time computer vision workloads?
This thesis compares a private 5G network (Ericsson infrastructure at Telefonica Germany) with enterprise Wi-Fi using a controlled video-streaming benchmark. A separate demonstration used an autonomous forklift in TUM's industrial testing hall.
The System
A camera on the forklift captures live video, streams it via RTMP to a media server (Oven Media Engine), which converts it to WebRTC for sub-second delivery. The browser-based interface receives the stream and runs object detection locally, eliminating a server round-trip. Five detection models were tested: COCO-SSD, YOLOv5s, YOLOv8s, YOLOv8x, and YOLOv9c.
Phase one: a controlled benchmark. 232 videos streamed through both networks, processed by all five models, producing 63,780 data points over 63 hours and 57 minutes. Every frame measured for latency, jitter, throughput, packet loss, and detection accuracy.
Phase two: a real-world demonstration. An autonomous forklift with a 360-degree camera at TUM's testing hall, remotely operated from the O2 Tower (Telefonica HQ) using a VR headset and Xbox controller, with live object detection over private 5G.
The Benchmark
Scale: 232 videos from the YouTube-BoundingBoxes dataset, combined into 6 hours and 23 minutes of 1080p footage. Each video streamed through both 5G and Wi-Fi, analyzed by five detection models. 31,920 Wi-Fi data points and 31,860 5G data points across 21 object classes.
Infrastructure: Ericsson private 5G (Radio Dot 4479, IRU 8846, Router 6675) at two locations: the O2 Tower and TUM's testing hall. Samsung Galaxy S21 as streaming source. Metrics captured per frame: jitter, latency, throughput, packet loss, IoU, confidence score, and inference time.
Models: COCO-SSD, YOLOv5s, YOLOv8s, YOLOv8x, and YOLOv9c, ranging from lightweight to state-of-the-art. All running in-browser via TensorFlow.js and ONNX Runtime Web.
Results
The benchmark reported similar mean detection accuracy: IoU of 0.646 (Wi-Fi) and 0.645 (5G), with no statistically significant difference (p > 0.05). This result alone does not establish statistical equivalence. No packet loss was recorded on either network during the test; the reported latency patterns differed.
| Metric | Wi-Fi | 5G |
|---|---|---|
| Mean Latency | 3.6 ms | 13.5 ms |
| Max Latency | 160.5 ms | 29.5 ms |
| Mean Jitter | 9.0 ms | 15.7 ms |
| Max Jitter | 53.0 ms | 27.0 ms |
| Peak Throughput | 12.9 Mbps | 23.5 Mbps |
| Packet Loss | 0.0% | 0.0% |
| Mean IoU | 0.646 | 0.645 |
| Mean Inference Time | 31.7 ms | 31.7 ms |
Network comparison across 63,780 data points. Highlighted rows show where 5G outperforms.
Latency consistency
Detection model performance (both networks combined)
Key Findings
- In this benchmark, private 5G and Wi-Fi had similar mean detection accuracy, with no statistically significant difference reported. The 5G latency measurements stayed within a 22 ms range, while Wi-Fi measurements reached over 160 ms.
- Wi-Fi had lower mean latency (3.6 vs 13.5 ms), while 5G had a lower recorded maximum (29.5 vs 160.5 ms). The averages and observed extremes describe different aspects of network performance.
- Peak throughput was 23.5 Mbps on 5G and 12.9 Mbps on Wi-Fi in this benchmark.
- No packet loss was recorded on either network during the benchmark. This observation is specific to the tested infrastructure and conditions.
- YOLOv5s had the highest reported F1 score among the tested models (0.858), compared with 0.849 for YOLOv8x. These rankings apply to the models and setup used in this benchmark.
- Object detection ran in the browser, avoiding a separate round-trip to a detection server. The public summary does not report a comparison with server-side detection.
The reported analysis found a positive association between throughput and detection accuracy (p = 0.0025), and a negative association between latency and confidence scores (p = 0.0005). No statistically significant association of latency or jitter with IoU was reported.
Brief
A controlled benchmark compared private 5G and enterprise Wi-Fi for video streaming and object detection. It streamed 232 videos through both networks, evaluated five detection models, and collected 63,780 data points. An autonomous forklift was used in a separate demonstration.
Wi-Fi had lower mean latency (3.6 versus 13.5 ms), while 5G had a lower recorded maximum (29.5 versus 160.5 ms). Mean IoU was 0.646 on Wi-Fi and 0.645 on 5G, with no statistically significant difference reported. That result does not establish statistical equivalence.
Key points
- 232 videos, five detection models, and 63,780 data points in the controlled benchmark.
- Wi-Fi had lower mean latency; 5G had a lower recorded maximum latency.
- Mean detection accuracy was similar, with no statistically significant difference. Equivalence was not established.
Takeaway
In this benchmark, Wi-Fi had lower average latency, while 5G had a lower recorded maximum. Mean detection accuracy was similar, with no statistically significant difference reported.