25
Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Embed Size (px)

Citation preview

Page 1: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Autonomous Vehicle

Instructor

Dr. Dongchul Kim

Presented ByHarish Kumar Gudipati

Page 2: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Introduction

Provides data signal information

Gives precise remote communication.

Can use this intelligence system for in

automating few tasks like Parking, Lane

changes and advanced driver assistance

systems.

Page 3: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Autonomous Valet Parking System(AVP)

Enables to drive and park without human assistance.

Earliest use is in Paris, France in 1905 and later developed in 1990’s and currently 25 AVPS projects are being developed .

Using mobile assistant system to select the parking lot and monitoring the state of vehicle in remote.

Page 4: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

how it works ?

https://www.youtube.com/watch?v=FYqe_Qy0OSo

Page 5: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Architecture of AVP System

AVP MOBILESelects parking lot

AVP SERVERGenerates drive path &

monitoring

AVP VEHICLE

CONTROLLER

Generates the parking path and follows it

Page 6: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

AVP Server System

Provides driving path for vehicle generated by considering vehicle

kinematics .

Driving path is generated as for reaching parking lot is main

objective as parking lot by verifying the occupancies too.

Detecting any obstacles through infra-sensor server and determines

the geometries of free space and additionally provides parking

area map named ParkGML to AVP mobile system for selecting final

parking lot.

Page 7: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

AVP Mobile System

User can make a request for AVP Service

It provides assistance to driver in mobile device.

State of vehicle can be monitored remotely.

User manually chooses the parking area in map as it

provides an user interface system with parking area

map

Page 8: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

AVP vehicle controller system

Controls vehicle position and vehicle movement by generating control commands – Steering speed and gear shift values to follow driving path.

To follow the driving path which is transmitted from AVP server in driving mode and next it generates list of control sections to control vehicle to park at final parking lot in Parking mode.

Detects the obstacles In the path and controls vehicle halt and resume forwarding in path by detecting obstacles

Page 9: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Parking Area Map Data Model (PARKGML)

It is a GML ( Geography Markup Language) established as a standard in ISO TC211 and OGC ( Open Geospatial Consortium).

It is modeling language for geographic systems and open interchange format for geographic transactions on internet.

It is generated using geometrical data in AVP server system for the purpose of monitoring background visualized data and selecting parking lot.

Creates memory resident data structure to generate vehicle path

Page 10: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

There are three kinds of data group as their specific properties.

Static feature is where the data will not be changed for long time which are generated by surveying and mapping.

Example : parking road and slot Semi dynamic features are data which will not be changed

during executing the AVP service after generated at the time of service started.

Example: Driving path and static obstacle date . Dynamic features : where the data will be continuously

changing during executing the AVP service in real time .

Example: Vehicle location and moving obstacles etc.

Page 11: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Park GML XML Schema

Page 12: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Park GML XML Schema Attributes

Have attributes like QueryType and ServiceType to implement the application protocol for AVP service.

Also has three profiles elements like driver , mobile device and vehicle profile in AVPUserProfile, assists in choosing the user-centric optimal parking lot.

Service attribute express the type of service and query attribute will request in AVP mobile system side to transmit ParkGML to AVP mobile system.

Page 13: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Driving Path Generation and following

Static grid-cell is used to implement path generation algorithm . It can be represented using the vehicle 4D kinematics(x, y, θ,r) to generate driving path. X and Y are center coordinates of grid cell. Θ is heading value of the vehicle and r means backward or forward driving.

Movable or immovable grid-cell is marked in the obstacle grid map.

1) ParkGML 2) Obstacle grid map 3) Voronoi grid map

Page 14: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Calculating Maximum Curvature angle

As the searching space is to be reduced and movable connected neighbor nodes are selected by maximum curvature angle of vehicle.

Maximum curvature angle is calculated by r value where r is minimum turning radius with steering angle.

Page 15: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Obstacle avoiding path A)The static obstacle avoiding path is generated by AVP serve system using

infra sensors and

B) we can see the backward turning path after forward driving in case of opposite direction to destination as below

Page 16: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Steering angle

It is important to know steering angle control value to correct vehicle lateral error while following driving path. It is calculated

using location/heading of vehicle as θsteering = (θ3 – θ2) / 2 -

θ1

Page 17: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Also the laser scanner sensors are used to avoid collision might be pedestrians or any vehicle etc. Used twoLMS151 of SICK corporation as a sensors.

Sensors are attached on front both side of a vehicle helps in measuring object distance maximum 40m within 0 ro 190 angular range . Obstacle avoidance path can be seen as follows green line is original path , blue is border line of original path and purple line is obstacle avoidance path.

Page 18: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Vehicle Parking algorithm Vehicle turning radius is minimum when steering angle is maximum where

θ steer = (θ t+ θ r)/2.

Adjusting is needed in θ  steer as steering angle control command value error, log points re gathered by control command value θ n ~ θ m . We can see error values by extracting steer angle value from circumscribed circle using log points as in below figure.

Page 19: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Backward perpendicular parking

The second minimum radius circle which meets at base point can be found at first. When driving the final path is sL the 1st minimum radius circle which meets at the line sL and 2nd minimum radius circle can be calculated

The control sequence of shift value is forward, backward and backward direction .

Vehicle parking mission is executing after driving change location of mission can be found using two minimum turning radius circles

Page 20: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Backward perpendicular parking control setup

After the parking path is able to generated consists three sub parts like start point to cross point, cross point to base point and base point to destination point.

Start point Cross point Base point Final point

Page 21: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Multi stage forward parking In forward parking more space is needed than backward and parallel parking.

It may not be parked just one stage at narrow space. Hence the multistage is used only in case of forward parking.

Backward moving path is generated and following before collision with other parking lots or other cars.

Figure a gives obstacles are detected by sonar sensors. Figure b gives the action repeated until final point is reached as shown in following diagram.

Page 22: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Multi stage forward parking algorithm

Page 23: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

DARPA Grand challenge

The implementation of this autonomous drive system has been implemented with little more innovation as the main goal is to develop the autonomous robot capable of traversing unhearsed off-road terrain.

It was developed by team of researchers to advance state-of-the-art in autonomous driving .

From broad perspective Stanley’s software mirrors common methodology in autonomous vehicle control .

DARPA challenge is a proof and milestone in quest for self-driving cars as it was tested on desert grounds and succeeded.

Page 24: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

Conclusion

The autonomous valet parking system provides the serviceusing digital parking map essentially.

As the future work, the technology to provide valet parking service is needed to be developed as real-time recognizing the parking line and land mark without digital map of parking area challengingly.

Page 25: Autonomous Vehicle Instructor Dr. Dongchul Kim Presented By Harish Kumar Gudipati

THANK YOU