Section Steel Alignment Correction Algorithm for Roller Conveyors Based on Double Deep Q-Network

The precise alignment of section steel on roller conveyor systems is essential for efficient transportation and processing, yet remains challenging due to the lack of accuracy of manual operation and the randomness of friction. To address the posture correction of section steel during the roller conveyor transport process, this paper proposes a posture correction algorithm for section steel on roller conveyors based on Double Deep Q-Network (DDQN). The objective is to maximize the waiting time before the baffle rises and minimize the time difference between the left and right endpoints of the section steel. First, a mathematical model is established based on the characteristics of section steel on the conveyor, analyzing the baffle operation time, the time difference after control, and the random effects of friction on transport. Then, two actions are designed for the baffle within the model: the waiting time and the control time. Corresponding reward functions are formulated according to the characteristics of each action. In the simulation environment, these two actions are integrated with the DDQN training method, and the training process is optimized. Finally, visual simulation demonstrates that the proposed DDQN-based algorithm can effectively adjust the posture of section steel. The evaluation metrics validate that the algorithm significantly improves posture correction accuracy, highlighting its potential as an effective solution.