Alexander Joos

Parallel Implementation of Constrained Nonlinear Model Predictive Controller for an FPGA-Based Onboard Flight Computer

Model Predictive Control (MPC) is an established control method in various application areas. Its ability of taking constraints into account makes it interesting also for automatic flight control. However, the computational complexity of MPC schemes usually limits its application. This paper describes a simple formulation of a constrained nonlinear MPC (NMPC) approach that can be realized on small onboard computers based on Field Programmable Gate Arrays (FPGAs). In contrary to classical implementations of MPCs a computationally expensive optimization problem can be avoided while even nonlinear prediction models and constraints can be considered. This is accomplished through parallel time-domain simulations. To this end, the parallel implementation properties of FPGAs are exploited. The 3d-kinematics is proposed as prediction model for the NMPC to plan the aircraft state trajectory (position and attitude) taking constraints and obstacles into account. Simulation results with a nonlinear 6 degree of freedom simulation model verify the functionality. Feasibility of hardware synthesis of parallel predicted models for the NMPC approach on an FPGA is shown by analysis.

Method for Parallel FPGA Implementation of Nonlinear Model Predictive Control

Abstract In this paper we describe a method for parallel real-time implementation of nonlinear model predictive control (NMPC) with constraints on a Field Programmable Gate Array (FPGA). A NMPC scheme based on time-domain simulations that we have introduced in earlier publications is briefly outlined. The parallel implementation on an FPGA is reported with special focus on computational resources, computation time, computational precision and FPGA-specific boundary conditions. Real-time testbed results in a 3-dimensional trajectory planning scenario for a fixed-wing unmanned aerial vehicle (UAV) with a FPGA-based onboard computer in the loop prove compliance of the scheme and the implementation.

Nonlinear Predictive Control based on Time-Domain Simulation for Automatic Landing

In this paper the problem of automatic landing of a fixed-wing aircraft with constrained nonlinear model predictive control (NMPC) is addressed. A parallel computable NMPC scheme based on time-domain optimization is used that allows for implementation on Field Programmable Gate Arrays (FPGAs). It is shown how a nonlinear 3d kinematics model with two control inputs (roll and pitch rate) can be used as a plant model. The control inputs are quantized and discretized and for this case controllability and stability is analyzed. An interpolating NMPC scheme is introduced that allows to reach precisely a commanded final position and attitude. In this scheme, flight mechanical constraints and obstacles are considered. This enables automatic landing with a commanded reference point, runway heading, glide path angle, and wings level. Simulations with a nonlinear 6 degrees of freedom simulation model show the applicability of the approach. Realtime testbed results demonstrate the parallel implementability on a low-power and low-weight FPGA based onboard computer.

Yaw Guidance for Airships Under Low Airspeed Conditions

*† This paper addresses the automatic tracking of a straight trajectory with an airship under the influence of wind and with low cruise velocity requirements. In this case any direct wind measurements are not available. The proposed new method is based on aligning the airship such that the body-fixed aerodynamic sideforce is minimized and thus, direct wind measurements can be avoided. The alignment can be realized through a new airship actuation with two lateral actuators that are used to create a yawing moment by powering them antithetically. At the low airspeeds considered here active yawing is not possible with aerodynamic control surfaces. However, using the aerodynamic sideforce as a control variable introduces a nonlinearity in the control loop. The closed loop stability is analyzed and the performance of this approach is demonstrated with a PC/104 flight control computer in real-time simulations under different conditions. A sophisticated nonlinear 6degree of freedom airship model with this new actuation approach is used as the demonstrator.

Control augmentation strategies for helicopters used as personal aerial vehicles in low-speed regime

In this paper an augmentation strategy is implemented with the goal of making the behavior of an actual helicopter similar to that of a new class of aerial systems called Personal Aerial Vehicles (PAVs). PAVs are meant to be flown by flight-naive pilots, i.e., pilots with minimal flight experience. One feature required for achieving this goal, is to have a Translation Rate Command (TRC) response type in the hover and low-speed regime. In this paper, a TRC response type is obtained for a UH-60 helicopter simulation model in hover and low-speed regime through the implementation of nonlinear backstepping control. The responses of the rotorcraft with TRC response type are evaluated with the metrics defined in the Aeronautical Design Standard ADS-33E-PRF. Simulations show the efficiency of the control scheme in tracking the reference velocities and the achievement of the requirements to have level 1 Handling Qualities (HQ) for the TRC response type.

Quadrocopter Ground Effect Compensation with Sliding Mode Control

In this paper the problem of automatic landing of a quadrocopter is addressed with emphasis on ground effect, battery voltage dependency, and resulting thrust disturbance. Ground effect and thrust dependency on battery voltage are quantified with flight test measurements. A ground effect simulation model is presented and approved with flight test data. Compensation functions both for ground effect and battery voltage dependency are developed and validated. The presented control design approach is based on the sliding mode control (SMC) theory where thrust disturbances are considered in the control design. The SMC control with compensation functions is tested in simulation and flight tested and compared to simulation results and flight tests with a classical PID approach. The proposed design shows improvements over classical linear control and reduces the disturbances efficiently.