Human-in-the-Loop Performance Evaluation of an Adaptive Control Framework with Long Short-Term Memory Augmentation

This study investigates the human-in-the-loop performance of a novel control framework, where a Long Short-Term Memory (LSTM) network augments an adaptive neural network (ANN) controller. The method drastically improves the transient response compared to conventional approaches, especially in the presence of significant and rapid changes in the uncertainties. LSTM network, which uses the knowledge of input sequence dependencies, predicts and compensates for the deviation of the ANN controller from its ideal behavior. Although this control framework is shown to provide improved transients, its interactions with a human operator need to be analyzed to ensure a safe operation. In this study, first, a human pilot model is used to investigate the overall system's behavior and analyze the controller's performance for a reference tracking task. Then, human-in-the-loop experiments are conducted to analyze how the system responds in the presence of a real human operator in the loop.