Autonomous Systems Lab

Where autonomous driving meets physical reality

A 1:6 scale model city for developing, testing, and breaking autonomous driving algorithms — safely, cheaply, and with real physics.

Research Facility

An 8m × 8m indoor model city with motion capture, traffic lights, and multiple autonomous vehicles operating simultaneously.

Overhead view of the 8×8m model town

Safe Environment

Test algorithms in a physical 1:6 scale model town — intersections, parking spots, and multi-lane roads with 83.2 cm road width and realistic markings.

1:6 scale model car with RealSense camera

Flexible Data Gathering

Combine onboard vehicle perception (RealSense D435, YDLidar G2) with global OptiTrack motion capture at 30 Hz for ground truth positioning.

Driving simulator with triple monitors and steering wheel

Human-in-the-Loop

Connect a full multi-monitor driving simulator for human steering and intervention — bridging automated and supervised driving research.

8×8m
Track Footprint
1:6
Scale Factor
6 vehicles
Run up to 6 cars simultaneously

Model Car

Purpose-built 1:6 scale autonomous vehicles with full sensor stacks, edge AI compute, and Ackermann steering.

Custom built · 1/6 scale · 3D print

Dimensions

Exterior (L×W×H)80×40×40 cm
Wheelbase50 cm
Track Width30 cm
Mass~15 kg

Equipment

Drive MotorBLDC EZRUN
Motor ControllerEZRUN MAX8 G2
LiDARYDLidar G2
CameraRealSense D435
Compute UnitJetson Orin AGX 64GB

Performance

Max Speed10 m/s
Max Steering Angle30°
Control InterfaceROS Ackermann
Speed ControlPID (rear diff RPS)

LUCCA Vehicle

Volkswagen eGolf · full-size · 5G connected

In addition to model-scale testing, the lab operates a full-size instrumented vehicle (eGolf). With an official autonomous driving license, LUCCA can be used for testing AD algorithms on public roads, full teleoperation, or large-scale data collection campaigns.

LiDAR

Reference LiDAROuster (roof)
Corner LiDAR4× Scala 2

Cameras

Teleoperation3× Cameras
Autonomous Driving4× Cameras

What We Investigate

Our pillars of research — all validated in the physical safety of our model town.

01

Sim2Real Transfer

Closing the gap between high-performing simulations and real-world application. We use ground truth injection to bypass onboard sensors and test with "perfect" perception data, then progressively add real-world noise.

02

Adversarial Safety

Verifying the adversarial robustness of autonomous systems in real-life scenarios. From physical adversarial patches to sensor attacks — testing resilience where it matters.

03

Planner Validation

Moving beyond synthetic tests and open-loop replays to validate planning methods in practice. Real vehicles, real intersections, real edge cases.

Collaborate With Us

We're actively seeking partners from industry and academia — for joint research, sponsorship, or anything in between.