In a small lab on the campus of the Egypt University of Informatics (EUI), six final-year students are quietly challenging the future of mobility. While global tech giants race to perfect autonomous vehicles, these young Egyptian innovators made the car could explain itself.
DriveFusion, a new AI-powered, transparent self-driving system that not only performs real-time driving decisions but also communicates the “why” behind every move.
Developed as a capstone project, DriveFusion marks a bold rethinking of how autonomous vehicles interact with human passengers—and a significant step toward explainable artificial intelligence (XAI) in safety-critical domains.
Multimodal AI Architecture
At the heart of DriveFusion is a multimodal AI architecture built on Qwen2.5-VL, an advanced vision-language model. Unlike most current autonomous systems—black boxes that leave even engineers guessing how decisions are made—DriveFusion fuses visual, spatial, and linguistic inputs into a unified decision-making system.
The platform integrates:
- Live camera feeds
- GPS and speed data
- User-initiated text queries
It then generates low-level control commands (like steering and braking), as well as high-level, human-readable explanations. A passenger could ask, “Why did you slow down?”—and the car would respond with an answer like, “There’s a pedestrian approaching the crosswalk.”
This combination of performance and explainability offers a level of accountability and user comfort that most commercial systems currently lack.
Ethics of Automation Redesigned
In a world where algorithmic opacity increasingly defines our relationship with machines, DriveFusion is a rare outlier—an AI system will be interpretable.
Sobhi, along with Eng. Toka Mohamed, a teaching assistant at EUI, helped guide the student team—Ibrahim Ahmed, Youssef Walid, Ahmed Walid, Saged Samer, and Mahmoud Khaled, alongside Samir—as they built the project over several months.
Their work directly confronts the two most persistent barriers to mainstream adoption of self-driving technology:
- Adaptability to real-world, unpredictable environments
- Transparency in decision-making and error accountability
An Egyptian Vision for Global Mobility
Though developed within an Egyptian university, the project speaks to global concerns—particularly the challenges regulators face in certifying self-driving cars that behave like black boxes.
DriveFusion’s design may offer a model for how emerging tech ecosystems in the Global South can lead with ethics-first innovation, especially in countries where skepticism toward automation remains high.
With DriveFusion, the EUI team is not only building a more intelligent car—they’re also building public confidence in a future where machines can be both powerful and principled.
