Real-World E2E Model Validation with BrightDrive
This is the first of several Autoware Community Demos we will highlight ahead of CES 2026, featuring member companies from across the ecosystem and their real-world Autoware deployments.
As part of this series, BrightDrive is showcasing how Autoware End-to-End (E2E) AI models can be validated on production-oriented autonomous vehicle platforms, operating in public road environments.
This demo will be presented at CES 2026 at the Autoware Community Demos Pop-Up Booth, located at the TIER IV booth (LVCC West Hall #6050).
BrightDrive — Deployment-Driven Autonomous Systems
– written by Alaa Mohamed and Mahmoud Serour (BrightDrive)
BrightDrive focuses on the development and deployment of Level 3 and Level 4 autonomous driving systems, spanning full-stack software, vehicle integration, and real-world validation across public-road environments.
Over the past years, BrightDrive has designed, integrated, and deployed autonomous vehicle platforms across Europe and the Middle East, covering both passenger transportation and logistics use cases. Its work consistently focuses on closing the gap between prototype-level autonomy and production-oriented, homologation-aware systems.
Key highlights include:
- Deployment of Level 4 autonomous buses in Dubai, Abu Dhabi, and Austria
- Development of the DriveAuto Stack, BrightDrive’s modular L3/L4 autonomous driving software platform
- End-to-end vehicle conversion to autonomous platforms, including:
- Sensor selection and field-of-view (FoV) layout design
- Multi-sensor calibration and synchronization
- HD mapping and localization pipelines
- Vehicle interface development (drive-by-wire integration and actuation control)
- Winner of the Dubai Self-Driving Competition 2023, demonstrating real-world system performance
Collaborating Through Open Validation
Within this community demo, BrightDrive focuses on validating and benchmarking selected Autoware components and E2E models under operational constraints, while sharing integration learnings and performance insights back with the Autoware community.
Areas of active validation include:
- Benchmarking selected Autoware components and reference models on BrightDrive platforms
- Validating Autoware E2E models, including AutoSpeed, SceneSeg, and DomainSeg, under real-world operational constraints
- Collaborating with the Privately Owned Vehicles (PoV) Working Group to deliver reference implementations grounded in operational data
- Sharing system-level performance metrics to support transparency and continuous improvement
This approach supports a shared vision of open autonomy, where innovation is accelerated through collective validation and real-world feedback.
Technical Synergy — DriveAuto × Autoware
BrightDrive’s DriveAuto Stack remains the primary production autonomy stack for commercial programs. In parallel, Autoware’s participation focuses on providing a fallback autonomy stack and a structured environment for benchmarking open-source components, without changing BrightDrive’s production stack direction.
BrightDrive’s L3/L4 zonal–domain architecture, powered by ROS 2 and NVIDIA-based compute platforms, is evolving into a next-generation platform known internally as QuantumDrive.
Key architectural features include:
- Hardware acceleration for real-time perception and decision-making workloads
- Adaptive AUTOSAR middleware for safety-critical and mixed-criticality applications
- GStreamer-based sensor plugins enabling deterministic multi-camera and LiDAR synchronization
- A modular, fail-operational system design supporting redundancy and remote supervision
This architecture enables production-oriented deployment while supporting structured benchmarking activities with the broader open-source ecosystem.
Autoware E2E Model Results — Public Roads
As part of this community demo, Autoware E2E models were validated on public roads in Slovenia.
Test configuration:
- Vehicle type: Van
- Accelerator: NVIDIA GeForce RTX 3050
- Pipeline stack: 10 FPS
- Models: AutoSpeed, SceneSeg
Results:
Stable real-time segmentation and object detection were achieved in mixed-traffic scenarios, demonstrating that Autoware E2E models can run reliably on production-oriented platforms in both urban and campus environments.
Roadmap and Looking Ahead
The BrightDrive–Autoware collaboration continues to evolve, with 2026 marking a key milestone in system robustness, validation depth, and deployment maturity.
Through community demos like this, Autoware aims to connect open research, industrial validation, and real-world mobility impact — accelerating open innovation across the autonomous driving ecosystem.
Learn More
About BrightDrive
BrightDrive develops autonomous technology for safer, more productive, and efficient mass transportation of people and goods. Based in Dubai, UAE, BrightDrive’s team brings world-class experience in developing automotive and autonomous driving systems for OEMs and Tier-1 suppliers globally.
About the Autoware Foundation
The Autoware Foundation is a non-profit organization supporting open-source software for autonomous driving — promoting safety, collaboration, and scalability across the global autonomy ecosystem.
