R&D Live Training Loop

Enabling real time messaging of raw sensor data and data processing with 5G.

5G enables online sensor data sharing from car fleets to a central server. This expedites the
traditional R&D pipeline used in Autonomous Driving, based essentially on collecting
datasets, training new machine learning models and testing these new models against
datasets. This pipeline is run in loop, until satisfactory results are obtained. Currently, the
datasets are stored locally in the car, and then moved to the laboratory in a hard disk. TIER1
and OEM car fleets are geographically distributed across the whole European continent, so the process can get quite slow. However, 5G enables sending the raw sensor data to a central server in real time. Not only this, but models can also be updated both in the cloud and in the car on demand, improving significantly the current process. This process optimisation can reduce the
already quite high R&D costs of CAM, and technical innovation can also be used for other
purposes such as forensics analysis to clarify product liability.


Before a new vehicle, autonomous driving or mobility system enters mass production, an OEM or a TIER1 performs field tests of these technologies. The Training Loop produced by the OEM or TIER1’s R&D department delivers different sensors to monitor driving dynamics, road conditions and system performance/ However, to optimise such system, new datasets are required to progress in the model train, test and update loop. When it comes to car-captured data, the volume is so high that storing that data in the vehicle for a later batch processing makes the process slow and complicated. 5G enables live messaging of raw sensor data for real-time processing and over- the-air software and models updates. The live and remote monitoring of the car systems allows real-time analysis and diagnosis to identify unseen conditions and expand datasets from the different dynamics captured by on-board sensors and cameras according to the driving dynamics.  R&D departments can optimise the new features and functions from data coming through a real-time pipeline.

5G enabled scenario

  • This use case needs the 5G eMBB feature, as the cars need to continuously upload data to the cloud service, which analyses new data to detect unseen conditions and trigger training processing to update machine learning/artificial intelligence/computer vision models to send a new update to all the testing fleets.

Years operating


European Countries