Real-Time 3D Occupancy Mapping by Multi-Agent Map Sharing using GPU Acceleration

In the next era of robotics, no machine will operate in isolation. Fleets of warehouse robots, swarms of autonomous drones, and agile robotic arms will work together—sharing one continuous, collective awareness of their environment. This PhD will advance perception and mapping by building a real-time 3D occupancy framework where multiple robotic agents exchange and fuse their spatial understanding into a single, living map. This shared awareness will not only power safety-critical responses like collision avoidance, but also feed AI systems that monitor and optimize the flow of goods, people, and resources across entire facilities. From split-second hazard detection to fine-tuned warehouse orchestration, this research will lay the foundation for truly collaborative, high-performance robotic ecosystems.

The research will focus on three key challenges:

1. Efficient Data Sharing: Creating robust distributed communication protocols (PROFIsafe over Wireless, private 5G etc. To be decided with the networking teams at imec) to synchronize occupancy maps between agents with minimal latency.

2.     High-Speed Mapping: Designing GPU-accelerated (CUDA) pipelines to process dense 3D data from multiple sensors in real time.

3.     Scalable Multi-Agent Coordination: Ensuring the framework performs reliably across heterogeneous platforms across different resolutions (ground robots, aerial drones, mobile manipulators).

The implementation will primarily be in C++ for maximum performance, with Python-based modules for rapid prototyping, integration with AI/ML models, and testing. The outcome will be a scalable, deployable mapping system that can operate in dynamic, unpredictable environments, paving the way for safer, smarter, and more collaborative applications.

Expected Expertise Domain of the Candidate

We are looking for a candidate with a strong interest in robotics, high-performance computing, and collaborative AI systems. Ideal candidates will have:
- Solid background in robotics, computer vision and communication systems
- Experience with CUDA programming and parallel computing
- Strong proficiency in C++ (with Python as a plus)
- Understanding of real-time systems and algorithm optimization
- Experience with dimensioning compute requirements and selecting the right platform
- (Preferred) Experience with 3D mapping frameworks (e.g., OctoMap, Voxblox) or multi-agent systems

Brief Description of the imec Research Group and Team

This PhD will be hosted at imec in collaboration with Vrije Universiteit Brussel (VUB). Imec is a world-leading R&D hub in nanoelectronics, AI, and high-performance computing, known for bridging breakthrough algorithms with cutting-edge hardware platforms. The hosting group specializes in autonomous robotics, embedded GPU acceleration, and real-time perception, working at the intersection of computer vision, AI, and system architecture.

You will join a multidisciplinary, international team with access to:
- State-of-the-art GPUs for CUDA development and testing
- Robotic platforms including mobile bases, aerial drones, and robotic manipulators with hands
- High-resolution 3D sensing equipment and motion capture systems
- Collaborative ties with leading academic and industrial partners in robotics and AI

The position is based in Brussels and Leuven, Belgium, with opportunities for collaboration visits to partner institutions and participation in international robotics competitions and conferences.

Supervisors: Bram Vanderborght, Constantin Scholz, Hamed Firouzipouyaei