Trajectory Tracking Control of Omnidirectional Robot Using Adaptive Fuzzy Control
avoin
Julkaisu on tekijänoikeussäännösten alainen. Teosta voi lukea ja tulostaa henkilökohtaista käyttöä varten. Käyttö kaupallisiin tarkoituksiin on kielletty.
Pysyvä osoite
Verkkojulkaisu
DOI
Tiivistelmä
Mecanum-wheeled robots can move in any direction and rotate at the same time. This makes them useful in tight industrial and service spaces. But getting them to follow a path accurately is not easy. Wheel–ground friction is nonlinear, the translational and rotational dynamics are coupled and the motor drivers have limited bandwidth. A simple fixed-gain controller can not handle these problems well.
In this thesis is presented an adaptive fuzzy gain-scheduled trajectory tracking controller for the Mecabot2, a four-wheeled mecanum robot running under ROS2. The controller works at the kinematic velocity level sending forward, lateral and angular velocity commands, which are converted to individual wheel speeds using mecanum inverse kinematics. A Mamdani fuzzy inference system replaces the fixed gains of a standard kinematic controller. The gains are computed online from the current tracking error and its rate of change, measured in the robot body frame. Three fuzzy systems run in parallel, one per correction channel. Lyapunov stability analysis shows the tracking errors stay bounded and converge asymptotically to zero.
Gazebo simulation tests on a circular path with correct parameterisation demonstrate accurate trajectory tracking performance, confirming that the controller converges reliably to the reference path. In this thesis, two common failure modes encountered during development: an initialisation error from wrong trajectory centre parameterisation and a phase-lag error from a logging timing mismatch between the odometry subscriber and the controller startup. Both failure modes are diagnosed, explained and corrected. Practical tuning guidelines come out of both findings. Hardware validation on the physical Mecabot2 platform is presented alongside the simulation results to assess real-world performance.