Autonomous formation flight of helicopters: model predictive control approach

Formation flight is the primary movement technique for teams of helicopters. However, the potential for accidents is greatly increased when helicopter teams are required to fly in tight formations and under harsh conditions. This dissertation proposes that the automation of helicopter formations is a realistic solution capable of alleviating risks. Helicopter formation flight operations in battlefield situations are highly dynamic and dangerous, and, therefore, we maintain that both a high-level formation management system and a distributed coordinated control algorithm should be implemented to help ensure safe formations. The starting point for safe autonomous formation flights is to design a distributed control law attenuating external disturbances coming into a formation, so that each vehicle can safely maintain sufficient clearance between it and all other vehicles. While conventional methods are limited to homogeneous formations, our decentralized model predictive control (MPC) approach allows for heterogeneity in a formation. In order to avoid the conservative nature inherent in distributed MPC algorithms, we begin by designing a stable MPC for individual vehicles, and then introducing carefully designed inter-agent coupling terms in a performance index. Thus the proposed algorithm works in a decentralized manner, and can be applied to the problem of helicopter formations comprised of heterogenous vehicles. Individual vehicles in a team may be confronted by various emerging situations that will require the capability for in-flight reconfiguration. We propose the concept of a formation manager to manage separation, join, and synchronization of flight course changes. The formation manager accepts an operator's commands, information from neighboring vehicles, and its own vehicle states. Inside the formation manager, there are multiple modes and complex mode switchings represented as a finite state machine (FSM). Based on the current mode and collected information, the FSM generates discrete reference points in state space. Then, the reference trajectory generator makes smooth trajectories from discrete reference points using interpolation and/or an online optimization scheme. By modifying the reference trajectory and triggering mode changes, the formation manager can override behaviors of the MPC controller. When a vehicle outside of the formation approaches a vehicle at the edge of the formation, the motion of the vehicle at the formation edge acts like a disturbance with respect to the vehicle attempting to join the formation. The vehicle at the edge of the formation cannot cooperate with any vehicle outside of the formation due to constraints on maintaining the existing formation. (Abstract shortened by UMI.)