We designed a nonlinear model predictive controller with collective blade pitch control for floating offshore wind turbines affected by strong winds. We derived a linear time-invariant wind turbine model from an operating point and constructed a nonlinear blade bending stress model for control design. Then, we designed a nonlinear model predictive controller by evaluating the stress model. The controller also includes disturbance previews of wind speed. Finally, we evaluated the control performances through numerical simulations on a precise numerical simulator, FAST. The results showed that the proposed controller reduced blade bending stresses more than other controllers.
This paper investigates a distributed management strategy for energy supply and demand networks. The strategy considered in this paper is the real-time pricing and decentralized decision making methodology, where the utility, an independent entity who want to realize the supply and demand balance, determines and supplies a conceptual price and each agent, which corresponds to energy storage unit in the problem considered here, determines own set-point of the active power injection as a solution to the individual optimization problem that includes the provided price. We see that the management strategy potentially involves issues of windup phenomena in the price provided by the utility, and we also see that the underlying mechanism causes windup is different from the one in the standard problem due to actuator saturation. The paper proposes a specific implementation procedure of the management strategy that overcomes the issue of windup.
This paper considers an application of a C/GMRES-based model predictive control (MPC) method to a diesel engine air path system. The plant model is derived based on the physical first principle to explicitly take account of plant nonlinearities. Since the plant has unmeasurable states, we employ an extended Kalman filter to estimate them. Then we design a C/GMRES-MPC algorithm and apply it to a real engine. We demonstrate the effectiveness of the present method by showing experimental results using a real vehicle.
This paper investigates blind spot coverage control for mobile visual sensor networks in a parking lot. First, a blind spot is defined as an area that is not captured by stationary monitoring cameras. Then, to cover the blind spot, mobile visual sensor agents receive the blind spot information from a server, and move to the target positions determined by their coverage rates. In this setting, we present a velocity-dependent Voronoi-based cut-in approach that guarantees global optimality and collision avoidance. Furthermore, the trade-off between transient optimality and convergence rate is analyzed. Finally, the effectiveness of this approach is demonstrated through numerical simulations.
This paper is concerned with a control problem for linear systems by means of pulse-width-modulation (PWM)-type input. Since the PWM input system is nonlinear in terms of the duty ratio, its analysis and design for precision control is in general difficult. Nevertheless it turns out that the discrete-time systems on the sampling instants can be exactly transformed to single input linear systems by making the number and center-locations of pulses variable, where the fact that the sets of the vector-values of the input terms are convex is used. In this paper, by utilizing the above fact further, we propose a method exactly transforming the nonlinear discrete-time systems to multi-input linear systems.
This paper proposes high response coverage control method of blind spots which cannot be captured by on-vehicle sensors and infrastructures. The conventional method is based on the Voronoi partition and gradient systems, and useful in intuitive perspicuity and formulation. However, the optimum solution depends on the initial displacement. This sometimes causes inefficient situations such that some fast agents cannot move to the other remote area with high risk potential beyond the own Voronoi partition. Therefore, in this paper, an additional motion control law is integrated into the conventional method considering the difference in speed capability of multi agents. The effectiveness of this approach is demonstrated using commercially available multi-copters.
In order to design control systems for ULVs (Ultra Lightweight Vehicles) with redundant actuators, a hierarchical control structure has been proposed. In the structure, the force distribution has been determined based on numerical optimizations. To determine the distribution, the inverse of the multivariate nonlinear function is required. This paper propose a method to compute an approximate but explicit inverse via proposing a method to construct the Lagrange inversion based on matrix computations. This is enabled by expanding the function as the power series in terms of the Kronecker products.
This paper is concerned with a stochastic model predictive control (SMPC) method for power management of a microgrid with large-scale photovoltaic (PV) energy supply. Recently, PV power generation has been increasing in total energy supply of electric power grids. For the management of microgrids, it is important to explicitly take account of the prediction of PV power generation because it heavily depends on the weather and hence is unstable. We propose a method for microgrid management by combining the SMPC and the PV power prediction. In our method, the PV power is predicted by Just-In-Time modeling from the weather forecast data. We demonstrate the effectiveness of the proposed method by numerical simulations.
As a technology of non-stop update without system restart, this paper proposes an automatic discriminant method which detects updatable parts and non-updatable parts of the program and an effect analysis method by program update. The novelty of this study is to apply control flow analysis based on Petri Net and structure analysis based on Kalman Decomposition to the program. By doing this, this paper aims to realize the judgment on the availability of update and the influence analysis of the update part while the program runs. Focusing on the program of robot lancer as an example of embedded systems, this paper reports results about the classification as updatable parts and non-updatable parts and the operation verification against the updated the program.
FRIT (Fictitious Reference Iterative Tuning) directly tunes control parameters from closed-loop input and output measurements without using mathematical models of the controlled plant. The pre-filter is used for initial input and output measurements for controller parameter tuning, and makes the data-driven cost function approximate to the original cost function as close as possible. This paper gives a design method for an optimal pre-filter in FRIT using closed-loop step response data by extending the method for VRFT. The paper shows that both the data-driven cost function and the original one have the same second order Taylor expansion with extended parameter descriptions. Finally, a numerical example shows the effectiveness of the proposed method.
Maintenance of the halo orbit using pulse input is considered for a spacecraft in the framework of the earth-moon elliptic restricted three-body problem. First, the dynamics around the orbit is expressed as a time-periodic discrete-time linear system. For this system, design of an optimal regulator is formulated into a time-periodic Riccati difference equation, which is solved backward in time. Once the solution converges to a periodic function, it gives an optimal regulator. The resulting controller depends on various parameters such as the input weight, the frequency of the pulse input, and the width of the pulse input. In order to have a controller with low fuel consumption, the control performance is evaluated with the total velocity change, which is closely related to the fuel consumption, and adjustment is made on those parameters.
A Gaussian process extended Kalman filter is effective for a state estimation problem when the nonlinear functions of systems are unknown. However, the Gaussian process extended Kalman filter is not adequate for judging some patterns where outliers are included in the observed values and states of the systems. This paper proposes an extended risk-sensitive filter, which is based on Gaussian process regression. The proposed method approximates the unknown nonlinear systems by using Gaussian process regression and estimates the states of the nonlinear systems with various outliers by using the extended risk-sensitive filter. Numerical simulations show the effectiveness of the proposed method.
Hydraulic excavators, which are used in some construction fields, are necessary for achieving energy savings and high productivity. Some excavators use a static compensator as the torque controller (referred to as a PQ controller) to maintain productivity. However, energy saving technologies complicate this system. Therefore, it may be difficult under certain conditions, such as low temperatures, to maintain both devices using the static compensator. As shown in previous studies, a PID controller with a double integral element has been used as the torque controller in the excavator (that is, it is used as the dynamic compensator). The hydraulic excavator modifies the system even if it performs only one motion. The system change is caused by changes in the pump pressure, which are based on the switching of hydraulic circuits. Therefore, changes in the internal state (changes in pressure) can be described as system events. In this paper, the event-driven torque controller of a hydraulic excavator is newly proposed. Moreover, the generalized minimum variance control method is applied to selected control parameters based on the closed-loop data for some conditions. Furthermore, the effectiveness of the proposed method is verified experimentally using the hydraulic excavator.
This paper gives a method of dynamic obstacle avoidance for a mobile robot with uncertain slips of the wheels. The control strategy is divided into two parts. The first part is the trajectory generation to avoid dynamic obstacles by solving a problem of mixed integer programming. The second part is the trajectory tracking of the mobile robot by tube-based model predictive control. Since the trajectory tracking error system includes a time-varying term due to the uncertain slips, this paper regards this term as disturbance and adopts the reachable sets with respect to the error to reduce the original constraints into tighter ones. As a result, the trajectory tracking error system does not violate the original constraints, which implies that the mobile robot can be expected to run with avoiding the obstacles. Numerical examples and experiments show the effectiveness of the method of dynamic obstacle avoidance for the mobile robot.
In this paper, we have developed an RT educational tool for physics in high school curriculum. Students, who use this tool, can see and feel physical phenomena and theories and they will consider questions by themselves. This system has enough sensors to make every physics experiment possible in a high school curriculum. We designed this sensors-system based on systematized physical quantities which we previously reported. We gave a class to four groups under different conditions. Group-A was with the RT tool, and group-B was traditional classroom learning, and Group-C was with the experiment tool as usual, and Group-D was without display system of the RT educational tool, to confirm visual effect of display. We carried out a questionnaire survey about each class, and tested students' comprehension on the class content. As a result, Group-A with the RT educational tool improved an understanding level and a result of the test together, and statistical significance was confirmed, although limited to a circular motion. Here we report the development and evaluation of RT educational tool for physics.
Recently increasing number of robots have been proposed for advanced patient care. In order to instruct robots that work in cluttered environments, we have introduced an intuitive interface based on a TOF laser sensor on a pan-tilt actuator. The interface enables us to control the direction of the laser spot to select an object and instruct a robot to manipulate it by dragging and dropping operation on a PC screen (“real-world click” interface). To manipulate objects everywhere in a room, the interface provides video images from a camera on the pan-tilt actuator (viewing window). However, the location of the laser spot is hard to see in the viewing window depending on the light condition or the color and the distance of objects. To cope with the problem, we propose a new interface that provides a marker shown in the viewing window at the position of the laser spot. In this paper, we propose a method to determine the location of the laser spot based on the distance to the target and camera parameters. An experiment on five subjects clearly showed the statistically significant improvement of the usability of the proposed marker.
In the steel industry, the role of the slab yard which is the intermediate process between the steel-making and the rolling is becoming significant. By promoting the logistics from steel-making to rolling, we can reduce the cost of fuel to reheating and improve the productivity. In the yard slabs are expected to be sorted into stacks as little transport as possible and as high as possible. It is the slab stacking problem. In this paper, we find the slab stacking problem has the similar structure with the set partitioning problem. And by using this feature, we formulate the slab stacking problem with the 0-1 programming. In addition, we show the efficient method of numerating the feasible stacks. Finally we prove the efficiency of this method with the computational experiment using the real operational data.