Trajectory optimization and control for autonomous helicopter shipboard landing

Thumbnail Image
Zhao, Di
Issue Date
Electronic thesis
Aeronautical engineering
Research Projects
Organizational Units
Journal Issue
Alternative Title
Helicopter shipboard landing presents a particularly challenging control problem because of (1) the limited landing time; (2) stringent safety constraints; (3) the turbulent shipboard motion resulted from the rough sea condition and (4) complex ship-airwake-helicopter interactions during the landing maneuver. While most shipboard landing operations are still carried out by human pilots, there is a critical need for the development of computer-assisted and/or fully autonomous landing strategies to relieve the pilot from significant workload and improve the operational safety and efficiency. Typically, autonomous flights are governed by the guidance, navigation and control (GNC) system, which replaces the role of the pilot in decision making, perception and action respectively. The GNC system comprises of hardware (e.g. sensors and actuators) and software (e.g. estimation, fault detection and control algorithms) components. This thesis concentrates on the development and validation of the guidance and control algorithms for landing a full-scale helicopter onto a ship deck in realistic maritime environments. For the purpose of autonomy algorithm design, a simplified nonlinear model has been developed to capture key characteristics of the full-state helicopter dynamics which are stabilized by a conventional dynamic inversion controller. Further, this simplified nonlinear model is shown to be differential flat, which enables the transformation of the original nonlinear system dynamics into an equivalent linear form with endogenous nonlinear constraints. For guidance, a trajectory optimization problem is formulated to achieve time-optimal landing, while complying with the simplified nonlinear dynamics and other operation constraints. Subsequently, the differential flatness property of this simplified dynamics is leveraged so that the original problem can be reformulated into a computationally-efficient form, where the handling of the free end-time end-state problem is streamlined by propagating the state trajectory in the linear flat output space. Furthermore, other than acquiring the numerical solution from the optimization problem using conventional temporal discretization technique, a basis parametrization technique is developed to further enhance the algorithm's computational performance. For outer loop trajectory tracking (horizontal position and velocity of the helicopter relative to the ship), a differential flatness-based outer loop controller has been designed to govern and stabilize the unstable zero dynamics. By canceling the nonlinearity using the endogenous mapping, the resultant error dynamics are rewritten as a linear fractional transformation. Based on this, robust stability and performance criteria are provided with respect to inner loop tracking performance, ship motion uncertainty and external disturbance. The effectiveness of this controller is then validated on the full-state nonlinear simulation platform. To manage landing under significant ship motion forecast error (in the heave direction), a reachability-based guidance and control method has been proposed. The key behavior of the vertical dynamics is encapsulated into the joint helicopter-ship dynamics, which consists of a nominal part and an error part. Accordingly, the reachable set and probabilistic reachable set are employed respectively to quantify the nominal boundary and actual distribution of the helicopter state relative the ship based on the knowledge of the dynamic disturbance and forecast error. Consequently, a shrinking horizon model predictive control strategy is designed to (1) minimize the expected terminal error of the relative position and velocity between the helicopter and the ship by choosing the optimal touchdown time, and (2) adjust the landing maneuver with regard to the updated ship motion forecast through recursive implementation. The capability of this strategy is then tested on the full-state nonlinear simulation platform.
August 2021
School of Engineering
Full Citation
Rensselaer Polytechnic Institute, Troy, NY
Terms of Use
PubMed ID