Enhancing UAV Altitude Control: PID Tuning with Reinforcement Learning and Genetic Algorithm

This paper explores the optimization of Proportional-Integral-Derivative (PID) controller gains to enhance the altitude control of Unmanned Aerial Vehicle (UAV)s, focusing on achieving precise tracking of set points and trajectories. Traditional manual tuning methods often yield suboptimal performance, limiting the controller's ability to balance stability and responsiveness under varying conditions. To address this limitation, we investigate automated optimization techniques, employing Reinforcement Learning (RL) and Genetic algorithm (GA) to refine the PID gains systematically. These methods enable dynamic solution space exploration, identifying parameter combinations that maximize controller performance. Comparative analyses reveal that algorithmically optimized gains consistently outperform those manually tuned by human operators. The results show significant improvements in stability and responsiveness. Using gains tuned by the RL, the system achieves altitude control with minimum overshoot. Integrating advanced optimization approaches into PID tuning enhances controller efficiency and significantly reduces the time and effort required for calibration. Experimental validation demonstrates the practical potential of these techniques, underscoring their effectiveness in improving the reliability and precision of aerial vehicles. This study emphasizes the role of intelligent optimization in advancing UAV control systems, paving the way for broader applications in autonomous aerial navigation and control systems. The video of the experiment conducted at Ericsson's 5G Autonomous Vehicles Lab, located at Carleton University, can be found here: https://youtu.be/oH3rQRhPMGs