Implementation of ANN on Microcontrollers in Line Follower Robots

Line follower robots commonly use conventional control methods such as if–else logic and PID control, which suffer from limited flexibility, poor adaptability to changing track conditions, and require complex parameter tuning. This study proposes a machine learning–based control system using an Artificial Neural Network (ANN) with a feedforward architecture and two hidden layers. The ANN model is trained using supervised learning with sensor data and motor output obtained from initial simulations, then implemented on an Arduino microcontroller for real-time control. Experimental results under both bright and completely dark lighting conditions show that the ANN-based system can control the robot stably and responsively, achieving a 90% success rate in completing the test track. These results demonstrate that ANN improves navigation accuracy, motion consistency, and adaptability, while also being effectively deployable on microcontrollers with limited computational resources.