The master's thesis addresses the problem of accurate and reliable control of a small autonomous vehicle in a confined indoor environment for the purpose of autonomously performing a spin jump. Performing a spin jump requires high speeds and rapid changes in direction, leaving only short time intervals available for motion correction. Such conditions pose a particular challenge for classical localization and control approaches, as relative methods such as odometry quickly accumulate errors, while absolute vision-based methods are sensitive to disturbances, delays, and limited fields of view. The aim of the work was to develop a system that enables robust, repeatable, and accurate autonomous driving of a remote-controlled vehicle towards a jump ramp and its automatic return to the starting point in a closed loop.
This thesis presents the design, implementation, and validation of an autonomous control system for a remote-controlled vehicle. The system fuses odometry, gyroscope measurements, and absolute ArUco marker observations using an Extended Kalman Filter, enabling robust state estimation even under fast motion and varying visibility conditions. A nonlinear kinematic model based on Ackermann steering mechanism is formulated, with the circular arc approximation used for state prediction. An efficient vision pipeline with dynamic region-of-interest adaptation is developed to reduce computational load and significantly increase detection frequency.
Vehicle control is achieved through a nonlinear orientation regulation method with an intermediate heading, ensuring smooth convergence to the target position while achieving the desired final orientation. Special operational modes are addressed, including continuous waypoint following for loop traversal and dynamic acceleration control for the launch maneuver. The entire system is implemented in ROS~2 and evaluated both in simulation and on a real remote-controlled platform. Experimental results demonstrate precise localization, stable control behavior, and repeatable execution of dynamic maneuvers, confirming the suitability of the proposed approach for advanced autonomous mobile systems.
|
