Item

Abstract

Control of a helicopter is a highly demanding task for the human operator, due to its unstable and coupled dynamics. Indeed, the pilot is required to constantly give inputs on the control device that are necessary to both move the vehicle towards a specific direction and, at the same time, to stabilize the system dynamics. Haptic support systems may be used as an alternative solution to help pilots in such demanding task. Design of an effective haptic system requires knowledge of the target trajectory. However, in many realistic scenarios, the target trajectory is not known in advance. For instance, in a helicopter free-flight task the pilot is free to choose any possible maneuver at any time, and the pilot intended trajectory cannot be known a priori. To provide the pilot with a haptic feedback that helps him/her to accomplish the intended maneuver, estimation of pilot intended trajectory is crucial. In this paper, a neural network approach is proposed to infer pilot intent based on data collected from maneuvers performed by an expert helicopter pilot, in a 2-DoF lateral/longitudinal scenario. Successively, a haptic feedback is designed to help the pilots to accomplish the intended trajectories. The proposed shared control system is evaluated in a human-in-the-loop experiment with minimally-trained participants in a fixed-base simulator. The participants performed a flight control task which included diagonal, lateral and longitudinal motions. Each participant performed the maneuver in two different conditions: with and without haptic feedback. Results showed effectiveness of the haptic feedback on participants performance compared to manual control.