The research behind real-world perception.
We take on the hard problems that decide whether perception holds up outside the lab. The work runs from simulation and synthetic data to formal guarantees, built hands on with students and graduates.
What we work on. The problems we keep pushing on so perception holds up in the field.
Closing the sim-to-real gap
Domain randomization and high-fidelity sensor simulation so models trained in simulation hold up in the field.
Synthetic data generation
Procedural scenes and structured datasets with pixel-perfect ground truth, from scene to sensor.
Formal verification
Provable guarantees for perception running across distributed robotic networks.
Human-in-the-loop
Keeping expert judgment inside the loop that produces production-grade perception models.
Active projects
- 01 Robust object detection trained on synthetic data
- 02 Modeling environments from publicly available maps
- 03 Data sharing and commercialisation
- 04 Simulation and data generation in Blender, Unity, and Unreal Engine
- 05 Intelligent assistants and LLM-based tools
Research collaboration
Joint projects, thesis work, and partnerships with labs and industry teams working on perception. Bring an idea or take one of ours.