Scoping the task, architecture and software, coordinating manufacturers and partners through to handover. Master’s degree in mechatronics and robotics.
Talk in the cybersecurity track: how machine learning protects fintech companies against deepfake fraud.
Excerpt from the FutureTech Con programme in Dubai, 9 May 2024.
THE TEAM BEHIND THE PROJECTS
One contact. More than 30 specialists behind it.
More than 30 specialists work on our projects, employed and freelance. For each project we put together a fixed core team. Heinrich Schneider leads it and remains your contact through to handover.
01
Software & interfaces
Web, apps and enterprise software, connections to CRM, ERP and phone systems.
02
Computer vision & voice AI
Video analytics on existing cameras, image models, phone assistants and call analysis.
03
Robotics & electronics
Mobile robots, robot arms, control and custom hardware.
SINCE 2017
From the lab into operation.
We started with research on robots, cameras and drones. Today our systems run in plants, kitchens, pickup points and estate agencies. Since 2017 we have completed over 100 projects for clients and our own products; many are under non-disclosure agreements, so here we show the ones we may talk about.
Electronic locks on a central controller, opening via app, time-limited access for couriers.
RESEARCH & DEVELOPMENT
Simulate first, then fly.
In this R&D project, a quadcopter learns to avoid obstacles on its own. The controller learned through reinforcement learning in a simulated environment. A neural network estimates the distance to obstacles from the image of a single camera, without a stereo camera or lidar. Tests with a real drone followed.
We apply the same principle in projects: where possible, we first test the logic on recordings or in a simulation and only then on the real equipment. That saves on-site testing time, and your operation keeps running as usual.
Test with the real drone in an underground car park. Top: the onboard camera image. Bottom: the depth map estimated from it (bright = near).
Indoor flight-test hall: lit gates serve as obstacles and fly-through targets.Simulation: the learned flight path (blue) avoids the blocks. Dashed: the direct line to the target.“Forest” training environment in the simulation. The network learns in changing scenes before it runs on real hardware.
PROJECT LEAD & INTEGRATION