Autonomous Cars for City Traffic Ral Rojas and the AutoNOMOS Team - PowerPoint PPT Presentation

Autonomous Cars for City Traffic Raúl Rojas and the AutoNOMOS Team Freie Universität Berlin Fachbereich Mathematik und Informatik Dahlem Center for Intelligent Systems

Topics Motivation The experimental vehicles Sensors Navigation and control Autonomous cars are green cars

The past of transportation The Great Horse-Manure Crisis In New York in 1900, the population of 100,000 horses produced 2.5 million pounds of horse manure per day

Automobile 1900

Green Cars: The End of Cheap Oil

The future of transportation

II The Vehicles

Team Berlin (FU Berlin)

Sensors in „Spirit of Berlin“

Sensors

MadeInGermany

MIG Sensors

Sensors in MadeInGermany 15

e-INSTEIN: electric & intelligent

Electronics in the trunk

iPad Remote Control

iPad Features

III Sensors  GPS and IMU Navigation  Laser scanners – two dimensional  3D Laser scanner  Video Cameras

GPS Positioning  IMU: Inertial Measurement Unit generates a true representation of vehicle motion in all three axes

Differential GPS in Germany

Ibeo Laser Scanner 200 meter range

Velodyne

One important city feature: trees, poles

III Computer Vision

Interior / Front Cameras

Automatic Sensor Calibration

Object recognition AdaBoost Stereo Traffic lights (Video)

Traffic Lights are Recognized

Control  Simulator for vehicle dynamics  High-level navigation planner  Low-level reactive control

Traffic rules are followed

Pedestrians are Detected

Architecture

Route Network Definition File (RNDF) ● Street Segments – GPS-Points – Lane width – Maximum speed ● Crossings – Entry-Exit pairs – Stop signs – Traffic lights 35

AutoNOMOS – Macroplan (KN) ● Macroplan: BFS in a graph ● Limited depth of plan (150 m) ● Microplan for each segment 36

AutoNOMOS – Microplan I maxShiftDist lateral shifts central shift / middle shift swerveDist swerveDist nudges swerves 37

Microplan - Transformation Transformation 38 38 Apr 11, 2012

Microplan (KN) ● Next few meters • Attached to lane spline • Evaluation function selects (50m) 39

AutoNOMOS – Evaluation function Weighted sum of: – non-collision [0,1] – Distance to checkpoint – time (t =s/v) – Curvature (1/r) – Maneuver value – heuristics 40

GUI – Controller/Behaviour 41

Parking and maneuvering Prof. Dr. Raúl Rojas, Freie Universität Berlin, Institut für Informatik - "AutoNOMOS" 42

Experiments in Traffic

Autonomous Wheelchair 2 x Computer (Linux & Windows) Kinect Ethernet 2 x Odometer CAN-Bus 2 x Lidar

Control Brain-Computer-Interface Eye tracking iPad iPhone

Sensor data Laserscanner(front & rear): Distances (8cm above the ground) Odometer (left & right): rotary encoders measure travelled distances . Kinect : RGB- and depth image.

Evolution vs Revolution: Driver Assistance Systems bester-fahrer.de

Partial Autonomy Prof. Dr. Raúl Rojas, Freie Universität Berlin, Institut für Informatik - "AutoNOMOS" 53