Robust Autonomy
for Certifiably Safe
Mobile Robots

Vision

Autonomous Mobile Robots (AMRs) could be mainstream.
They could deliver the mail. They could plow our fields.
They could extinguish forest fires.Building a robot could be as easy as building a video game: just focus on the content, take building blocks from a high-quality game engine. We want to make that happen.

Top-down view: a robot stops short of a safety radius around a person

The AI robot software stack

We know that the middle layer of the robotic stack is crucial.
It has to be of high quality. And it has to be designed and
certified for functional safety.

Application Layer

Autonomy
Middle Layer

AMR / AGV Platform

interfaces to orchestration & application logic

Robust Perception

Segmentation, Object Detection, Traversability

Reliable Localization

novel robust and real-time sensor fusion

Safe Guidance

Fail-safe trajectory generation

interfaces to low-level control & sensors

Localization

At any time, a robot needs to be able to enter a safe state. That means safely decelerating, avoiding obstacles and coming to stand-still at a safe spot.That's why the first component we tackled is localization: Without it, no controlled move can be taken. That's why currently nearly all mobile robots rely on a simple proximity based emergency brake system - far from true autonomy.

Top-down view: a robot follows a safe path around a person

Gneiss for localization

To design smart autonomous behavior, relative localization
is crucial. Failure might be fatal. We solve it with our proprietary sensor-agnostic sensor fusion gneiss. It is deterministic and has limited error bounds, so it's certifiable.