Trajectory optimization and stabilization in the control of pneumatically actuated drones

As humans prepare for lunar and deep-space missions, autonomous systems for atmosphere- lacking environments are essential. This paper advances control methodologies for a pneumatically actuated 2D hopper. We develop and compare delayed hysteresis, PID with anti- windup, bang-bang control, MPC, and reinforcement learning (DDPG) combined with Kalman filtering for state estimation. Using LiDAR and MPU6050 sensor data, the system improves responsiveness through accelerometer feedback. Experimental results show model-informed observation strategies outperform traditional methods, offering a robust platform for testing autonomous control in conditions analogous to vacuum environments.