11
CONFIDENTIAL ©2014 GlobalLogic Inc. DIY Self-Driving Car Overview an HowTos

DIY Self-Driving Car

Embed Size (px)

Citation preview

Page 1: DIY Self-Driving Car

CONFIDENTIAL©2014 GlobalLogic Inc.

DIY Self-Driving Car

Overview an HowTos

Page 2: DIY Self-Driving Car

2 CONFIDENTIAL

The beginnings

1950 Smart Highwayswith steel cable in the roadGM and RCA, USA

1977 Machine visiontracked white road markingTsukuba Mechanical Engineering Lab, Japan

1997 PROMETHEUS 1200 mi, 94% autonomouslyARGO Project, Italy

2004 DARPA ContestUSA

Page 3: DIY Self-Driving Car

3 CONFIDENTIAL

DARPA: The challenge

− 2004 – less than 12 of 240 km route completed

− 2005 – 5 of 23 vehicles finished the course. 4 in 7.5 hrs

− The winner was Stanley (Stanford Uni)

− 2007 – Grand (Urban) Challenge – City model with traffic: 1st Tartan Racing (Chevy Tahoe, Carnegie Uni) – 14 mph;

2nd Stanford Racing (VW Passat, Stanford Uni) – 13.7 mph;

3rd – VictorTango (Ford Hybrid Escape, Virginia Polytechnic Inst) – 13 mph.

− 2013 – Fast Adaptable Next-Generation Ground Vehicle (FANG) Challenge – military platforms contest.

Page 4: DIY Self-Driving Car

4 CONFIDENTIAL

Google: Human drivers are the problemManned vehicle T-crossing VS unmanned

Advantages:

• No accidents

• Denser traffic

• Less stops – less jam

• Higher vehicle coordination

• Lower travel time

• Lower air pollution

Unmanned vehicle T-crossing

Page 5: DIY Self-Driving Car

5 CONFIDENTIAL

Self-Driving Car (SDC) – What’s inside?

− A lot of sensors – more is better: velocity, acceleration, magnetic sensors; gyroscope, compass, range sensors, video, LIDAR, GPS,

− Control – steering, speed/brakes, transmission, signaling lights, horn etc,

− Computing power – distributed computing (sensor’s controllers), centralized computing (multi-core, even up to 40 – MIT’s Talos),

− Connectivity – WWW, inter-SDC,

− UI

Page 6: DIY Self-Driving Car

6 CONFIDENTIAL

Spinal cord: Sensors and Controls init, coordination and

interaction, emergency handling

CtrlrS2 CtrlrSNCtrlrS1

Self-Driving Car – How it works?Sensor1

UI/AI: Goal definition, Route planning, Emergency

recovering

Sensor2 SensorN

Hard realtime

RealtimeBusiness logic (AI?):

Motion planning, Positioning, Emergency avoiding

CtrlrC1 Control1

CtrlrC2 Control2

CtrlrCN ControlN

Not realtime

Page 7: DIY Self-Driving Car

7 CONFIDENTIAL

Programming the SDCGetSensorsData();

If(Sensor1OK) do(actionOnSensor1(Sensor1Data));

If(Sensor2OK)

If(resActionOnSensor1 == res1onSensor1) do(action1OnSensor1(Sensor1Data))

else do(action2OnSensor1(Sensor1Data))

If(Sensor3OK)

If(state == Stopped)

If(resActionOnSensor2 == res1onSensor2)

If(…)

else …

Page 8: DIY Self-Driving Car

8 CONFIDENTIAL

Programming the SDC Cntd.GetSensorsData()

AnalyseSensorsData(CurrentState[])

MotionPlan()

CalculateControls()

Go()

CurrentState[]:

P1, C1,P2, C2, … PM, CM

Sensor1 10 .99 3 .857 .3

Sensor2 8 .7 3 .160 .9

SensorN 15 0 4 .955 .7

Page 9: DIY Self-Driving Car

9 CONFIDENTIAL

Legend

Route

Desired position

Sensors’ data

Obtained position

What we get

Page 10: DIY Self-Driving Car

10 CONFIDENTIAL

Our Big Goal

− Have our RoboCar running and avoiding obstacles,

− Teach it to create its world’s map and navigating in it,

− Persuade it to charge its batteries at Docking station when depleted,

− Make it to interact with surrounding,

− Do it all having a lot of fun.

Page 11: DIY Self-Driving Car

CONFIDENTIAL©2014 GlobalLogic Inc.

Thank you!