Dynamic hyperparameter tuning-based path tracking control for robotic rollers working on earth-rock dam under complex construction conditions

Abstract

Robotic roller technology is important for ensuring the compaction quality and construction efficiency of earth-rock dams, and its key challenge is related to path tracking control. The current easy-to-use and effective proportion–integration–differentiation (PID) path tracking controller requires time-consuming and labor-intensive hyperparameter tuning, making its application in complex construction environments with a set of fixed hyperparameters, such as transition, misalignment, and position drift, difficult. This paper describes the TSABFA-PID path tracking controller based on hybrid tunicate swarm algorithm (TSA) and bacterial foraging algorithm (BFA), and modified kinematic model. The proposed controller can dynamically self-tune hyperparameters to adapt to complex construction condition by optimizing the objective function. The proposed controller achieved better results in both simulation environment and on-site tests than state-of-the-art methods. In the future, the deep reinforcement learning and human-computer interaction techniques will be integrated into the proposed controller to improve the compaction performance.