Designing of omni directional, autonomous and ai controlled ball launcher

Abstract

This work began as a way to improve football training by providing steady, accurate passes without needing an extra person. The project designs an omnidirectional, AI-based ball launcher that combines mechanical hardware, aerodynamic flight mod eling, and real-time computer vision. A lightweight aluminum chassis and carefully selected motors keep the system mobile while maintaining safe operation. Drag, lift, and spin forces are included in the ball-flight model to support precise launches. Reinforcement learning methods such as Proximal Policy Optimization (PPO) and Deep Deterministic Policy Gradient (DDPG) are then used to adapt the aim based on observed outcomes. Early tests indicate that the launcher gradually improves its accuracy, delivering the ball closer to the intended target. By automating repetitive passing tasks, players can focus on skill development while training becomes more efficient and consistent.