Path tracking learning methods 1/2 - Matlab/CarMaker simulation
Path tracking with adaptive feedback gain - CarMaker simulation
Adaptation of feedback gain by sensitivity estimation
Self-tuning of adaptation gain by personalized rule
Generated path can be used for various scenarios
- (Input) Offset, length, width, point interval
- (Output) Waypoints
Adaptive path tracking at virtual HKNU - CarMaker simulation
Generated path from actual road condition at HKNU
Learning MPC-based path tracking - CarMaker simulation
Autonomous Mobility - Steering and Driving Control
Software : ROS
Function : path tracking, safety control
Controller : longitudinal and lateral control
- (goal) Kinematics-based desired acceleration
- (goal) Model-free steering controller
Software : Matlab/Simulink + CarMaker
Target motion : virtual car-following
Controller : Model-free adaptive feedback controller (MF-AFC)
: Using gradient-descent method
: Sliding mode control algorithm
Mass : about 300 kg
Size : about 1,600 mm x 1,160 mm x 550 mm
Velocity : ~ 20 kph (max. 40 kph)
Environment sensors for autonomous driving (to be installed later)
Lidar(Sick)
Camera(Microsoft webcam)
LDS lidar for object detection
Detection range : 10 ~ 3,500 mm
Field of view : 360 deg
Angular resolution : 1 deg
Scanning frequency : 5 Hz
Two wheel driving module
Motor : brushless DC motor
Input voltage : 24 V
Rear wheel : free rotational wheel w.r.t z-axis
Maximum speed(longitudinal) : 1.5 m/s
Fault reconstruction / detection / isolation
Safe / warning / emergency
Tolerant control / emergency parking