jcvikl / SRC

Self driving Rc Car

SRC robot (Self driving RC Car)

Fully hand-made metal mobile robot

Official website: https://yunbum.github.io/

As [Self Driving robot Engineer] I made fully hand-made metal Self Driving RC car.

Main UI

• HW setting: id, rate, type...
• SW setting: config file selection, detection range, vision ROI...

HW Detail

• Main Controller: Latte panda(Window or Ubuntu), Raspberry pi
• Sub Contoller: Arduino > motor, light, LED dot matrix control
• IMU (AHRS): Heading direction check
• Lidar: Emergency stop, collision avoidance
• Camera: Lane detection, Machine Learning (Object detection)
• Speaker: Connection beep, Alarm, Music(wav)
• Light: 12 LED for light / mode check, night driving test

GNSS/GPS

• GNSS: RTK fixed or float mode setting (RTCM message) using NTRIP client
• NTRIP client: u-center, Lefebure
• Base Station: gnss.eseoul.go.kr, vrs3.ngii.go.kr
• Map: Google satellite, Bing..

IMU/ AHRS

• AHRS Heading Kalman Filter

Camera / webcam

• Image process: Binary conversion, ROI setting, Clamp, Area calculation
• interface: PoE, USB with camera file (genicam)

Lidar

• Data processing: 3D, 2D raw data gathering, Clustering
• Comm: TCP/IP, UDP, Com depend on manufacture
• Range set: detectin range set > horizontal, vertical

LED light

• 12 Led module using relay switch module
• color: bule, white, yellow ...

SRC Type

Led dot matrix

• Display info: distance gap, degree gap, time, Latitude, Longitude,...
• Mode select: joystick
• Custom: Add any personal message or customized display text

Joystick

• SRC Manual control
• Mode select: Manual/ Self Driving
• Etc: Sound check, LED Light on/off

Camera /vision

• image processing: Binary conver

Speaker

• Function: connection alarm, Music,

SW Detail

Steering

• Steering: pure pursuit and PID control
• PID: minimize the angle difference between waypoint and agv-lookahead point

Waypoint

• GPX format: Using GPX route editor to create, modify the route easily.
• Basic: Lookahead point, Target position, Heading value > minimize degree gab(waypoint vs AGV-lookahead deg)
• Tools: Using GPX route editor, easy to make and modify waypoint for driving
• Coordination: WGS84 to TM transform

Log replay

• replay the logfile after driving.
• waypoint / actual driving route compare
• statistics: distance error max/min/average
• control value check: PID, stop point etc check..

Reinforcement Learning

• Model: I'm testing DQN logic to apply driving algorism