Test Solutions for Camera ECUs and Big Data at Cenex Expo in England

Visit Konrad Technologies Stand Located at C2-105
‍04. – 05. September 2024
UTAC Millbrook
Millbrook, Bedford, MK45 2JQ, England
‍https://cenex-expo.com/

Konrad Technologies to exhibit a range of ADAS/AD test technologies demonstrating development from the lab, into vehicles and subsequently the analysis of Big Data from fleet deployments.

The exhibition journey begins with an example of laboratory testing of vision-based Electronic Control Units (ECUs). This is followed by comprehensive in-vehicle data logging of these ECUs and the associated vehicle sensors. The collected data is then seamlessly transferred to data scientists through a robust data pipeline. Leveraging human-assisted AI from our partner Akridata, we then visually explore, curate, and identify the most valuable 1 % of images and videos from real fleet deployments, ensuring precise and actionable insights.

Presentations

Day 1:
Test Drive to Cloud: Building an End-to-End Data Workflow to Accelerate & Scale CAM Data Insights
‍Session "The Role of Data in enabling CAM" starts 14:45, talk 15:05
Speaker: Robin Irwin

In-vehicle data captured at petabyte scale from global test drives brings an array of complex challenges. This presentation explores new CAM tools, infrastructure and methods to accelerate learnings and put the“right” data in the hands of data scientists from vehicle (far-edge) to cloud across an end-to-end data pipeline approach.

Day 2:
Hardware-in-the-Loop for ADAS/AD Camera Validation‍
Session "Verification & Validation" starts 14:30, talk 14:50
Speaker: Robin Irwin

A presentation of HIL (hardware-in-the-loop) methodologies and tools for verifying and validating ADAS / AD engine control units (ECUs) where real-time images are directly injected into ECUs to emulate the real world.

Demonstrations

"CAMulator" Camera Direct Injection HIL

Automotive OEMs and suppliers must develop robust ADAS/AD systems based on vision cameras, hence must validate detection, perception and control algorithms within the camera ECUs. It’s crucial to test ECUs in a diverse range of scenarios and hence virtual simulation in laboratories is a common approach, injecting synthetic data generated by any number of simulation packages on the market. 

The Konrad Technologies CAMulator is a camera direct injection solution that serves as a data bridge from virtual simulators to physical ECUs, emulating real world cameras, proving out common features such as lane keep assist (LKA) or automated parking, as well as a tool for firmware & software regression testing. Offering compatibility with any simulation tool via an array of up to 16 HDMI inputs, CAMulator can configure and synchronise multiple camera image data with backchannel communications. A single UI can input from HDMI or recorded files, switch to real cameras in“TAP” mode, embed data manipulation (such as format conversion, custom frame rate, customised padding, frame drop, colour calibration) and provide GPIO synchronisation, all within a unified user interface for all settings and control.
 
Interactive CAM Workbench 

Each connected and autonomous vehicle generates TB of big data daily, and like any big data problem, the 5 V’s of big data (velocity, volume, value, variety and veracity) are the five main characteristics. On a fleet scale this presents major challenges in cost, complexity and control, with data volume is in PB per year. This presents a challenging data pipeline -from the far edge (the start of the journey in the vehicle), right the way to cloud infrastructure and finally in the hands of the data scientist. Konrad in-vehicle data loggers capture this raw, high BW, high data rate data from sensors and ECUs for subsequent analyse, but how should data scientists find the useful 1 % of data quickly and easily when data is so complex and challenging on this large scale?

Konrad Technologies Data Explorer is an interactive workbench for ADAS/AD, computer vision, AI engineers and data scientists,enabling exploration, search and analysis of images, videos and point clouds captured from CAM test drives. This workbench significantly reduces the time data science teams spend on data preparation, curation and selection of data, with a human assisted AI approach to data curation. Through its visually interactive and intuitive design, data science teams can quickly explore millions of raw images or videos, identify points of interest, perform visual search and build effective training data sets for labelling and training. An off-the-shelf workbench can reduce hours of work to minutes by cutting time spent on data selection and curation, avoid wastage in labelling spends using that method of meta enrichment, and accelerate your path to model accuracy, reducing the iterations and guesswork.

About Cenex Expo

Cenex Expo, originally Cenex-LCV and Cenex-CAM, is the leading event in the UK innovation community. With over 4,000 visitors in attendance last year, the event offers a wide range of exhibitors, an extensive seminar programme and many networking opportunities. As part of the outdoor exhibition there is also the opportunity to ride & drive the latest research & development vehicles. Supported by government agencies, the event brings together the people who will deliver the net zero and connected automated mobility future.