Search

In this paper, we propose an attitude control law for underactuated two-wheel spacecraft under non-zero total angular momentum. Attitude control with non-zero total angular momentum is complicated ...

Vision sensors are widely used not only for detection and recognition, but also measurement of the pose of the moving objects. But in the crowded environment, occlusion often disrupts the measurement, and wrong data association due to misrecognition deteriorates the tracking performance. Probabilistic data association filter (PDAF) is known as useful to address such issues, in which observed features are weighted by probability to deal with multiple observations as well as to cope with occlusion and false recognition. This paper presents a novel object tracking method in which PDAF is incorporated into moving horizon estimation (MHE) framework to deal with multiple frame tracking and physical constraints. The performance of the proposed method is evaluated by comparing with the PDAF.

Optical motion capture system can measure the position and the orientation of an object equipped with retroreflective markers. When some markers are not observed due to occlusion, the system often fails to localize the object. In our past study, moving horizon estimation (MHE) is applied to improve the estimation robustness, where states are determined so that the sum of error magnitude of dynamics and observation is minimized under constraints. In general, marker association is crucial for high accuracy estimation. In this paper, we propose a marker association method for MHE where the proposed association method exploits a potential function and optimization so that the association is robust against marker occlusion and fake images similar to the marker. In the simulation, for the environment where the previous method fails due to missing markers and misrecognition, we confirm that the proposed method can successfully realize the estimation.

This paper presents energy conserving vehicle motion control for electric vehicles equipped with different types of drive motors. The proposed controller, which minimizes total energy consumption in a vehicle, consists of guidance control and distribution control. In the guidance controller, optimized vehicle motion conserving energy consumption is generated using model predictive control. In the distribution controller, dissipation energy is minimized utilizing a redundant set of tire forces. By separating guidance and distribution control, the complexity of vehicle dynamics which is considered in nonlinear MPC is reduced, and computational load is also reduced. Effectiveness of the proposed energy conserving control is shown through the numerical simulation using a test course modeled an interchange; the energy conserving vehicle behavior is shown.

Leg/wheel robots has a hybrid movement mechanism which combines leg and wheel mechanism. In this paper, we propose a posture control for three-dimensional leg/wheel robots in which the distance between zero moment point and center of gravity are minimized using model predictive control. It is expected to prevent overturning of the robots based on prediction and considering zero moment point. The proposed method can be applied to the robots with redundant degrees of freedom. We show the superiority of the proposed method through simulations of multiple body dynamics.

When leg/wheel mobile robots are controlled by model predictive control (MPC), the optimization is too complex for an embedded CPU. Then, we proposed a method which reduces computational complexity by sequential optimization for the models partitioned into each leg. However, robustness of the proposed method was not evaluated enough in the disturbing environment. In this paper, we conduct simulations on the inclined plane to evaluate the robustness. Through the simulations, the robustness of the proposed method is improved by changing the control cycle and the predictive length.

Localization with laser range finder (LRF) needs an environment map. In some environments, the localization using the environment map prepared in advance often fails. In this study, we achieve localization using both the environment map and the feature of supplemented objects as the estimated map. Moreover, to suppress the influence of the observation noises of LRF, we estimate the feature of supplemented objects using the extended Kalman filter (EKF). In this paper, we verify the influence of the estimated map on estimation accuracy through the simulation and verify the effectiveness of this localization method through the experiment conducted in a gymnasium where is larger than the scanning range of LRF.

This paper combines coverage control, random-walk and artificial potential method to deploy multi-copter type Unmanned Aerial Vehicles (UAVs). With the coverage control, it is able to deploy agents in the coverage area with arbitrary distribution based on the purpose of control. However, when initial position of agents concentrates on one location, it could be fallen into the local optimal solution (local optimal problem). While the local optimal problem, agents could not achieve the distribution based on the purpose of control. Meanwhile, random-walk takes vectorial angle randomly and move constant distance. By using this movement, it is possible to disperse the initial position of UAVs. Therefore, the local optimal problem could be solved by combining random-walk to coverage control. Moreover, using the artificial potential method as a repulsive force to avoid the collision between agents. This paper verifies the validity of proposed method by numerical simulation.

We apply the method to suppress the influence of the model error to a quadcopter. Because a quadcopter is a flying object, the stability is important. Many model based controllers have been proposed and applied to a quadcopter but the control performance is deteriorated when a model error exists. The model error compensator (MEC) is applied to a model based controller as a solution of this problem. MEC is one of the 2-DOF control system. In many studies using the model based control, a linearized model is used to simplify the controller design, but model errors may change by the attitude of a quadcopter. The compensating the fluctuated model error is difficult using simple MEC. Therefore, the fictitious reference iterative tuning (FRIT) is applied to adjust the compensation gain of MEC. using the optimization calculation. The performance of the proposed controlled system is verified by experiments.

This study presents the interaction suppression control of multi-copter using MPC that considers the disturbance map. The influence of wind by propellers to another quad copter is estimated beforehand and the disturbance map made from the information. The disturbance map stores the estimated disturbance value corresponding to a relative position. Also, the disturbance map is implemented in the control model for model predictive control. Hence, it is possible that the controller explicitly considers the future expected disturbance via a disturbance map. This paper performs the disturbance suppression control experiments with actual quad copters. and shows the effectiveness of the proposed method.

This paper newly proposes an on-line learning method for control maps by using Gaussian filters. In the present method, man-hours for calibration of control maps can be decreased, and complicated structures of control maps are learned without prior knowledge. Moreover, by the effect of Gaussian filtering, smoothed maps can be created even under noisy conditions or few measured points. We also introduce improvements of the algorithm to cope with engine deterioration due to aging. In this work, the proposed method is applied to minimum advance for best torque control on actual vehicles with just one driving data, and the accuracy of the learned map is verified through simulation and experiments.

In this paper, inter-vehicular distance control method is applied to the small vehicles such as automatic guided vehicles (AGVs). Path tracking and attitude angular error will turn up when vehicles take a detour to avoid obstacles, however, this phenomenon has not been explicitly considered in conventional methods. In this paper, we propose target velocity modification considering path tracking and attitude angular error and assess the effectiveness of it quantitatively by numerical simulation in a situation of obstacle avoidance.

In this paper, we apply the hierarchical model predictive tracking control for real robot and verify the travelling performance. We deal with the independent four-wheel driving/steering vehicle (IFWDS). This vehicle is equipped with the four coaxial steering mechanisms (CSM). It is composed of two steering joints which have different movable range and rotational speed. In our previous study, we achieved the improvement of tracking performance by optimal allocation of steering angle for CSM in simulation. In this study, we verify that we can obtain the same result in experiment.

Motion capture systems are widely used for measuring the position and the orientation of objects. It detects the markers attached to the objects, but they are sometimes lost sight by occlusion and dead angle of sensor. It becomes a problem that markers to determine the coordinates of the object are not measured. Thus, it is necessary to estimate state suppressing the influence of missing measurement. In this study, we propose a Moving Horizon Estimation for state estimation to tackle this issue. If any markers are not measured because of occlusion, MHE can estimate position of the robot uniquely because it optimizes the evaluation considering the error of available measurement and motion dynamics of the vehicle. In our previous study, we applied MHE for estimation of the position and heading angle of the vehicle robot with markers. In this paper, we will show the application of it to the visual feedback control for the robot using estimation value at real time. we will prove that MHE is effective because it can suppress the influence of occlusion and rapid change of input.

A safety technology about unmanned aerial vehicles(UAVs) attracts a lot of attention. This paper presents a robust formation control method for multi agent system (MAS) to suppress the influence of disturbance and maintain the formation robustly by sharing the disturbance information. This paper estimates the disturbance by the disturbance observer. UAVs share these disturbance information using network of MAS. The inputs are calculated by model predictive control(MPC) that predicts the future motion of each UAV considering a shared disturbance information. UAVs share the disturbance information and the future motion of UAVs predicted by MPC using network of MAS. The proposed method achieves the robust formation control to consider these future information in MPC. This paper verifies the validity of proposed method via simulation.

In our previous study, we proposed a localization method based on moving horizon estimation (MHE) for known environments. However, in the method the environment map is required for achieving the localization. So in this study, to achieve the localization in unknown environments, we propose a localization method integrating scan-matching and odometry based on MHE; the vehicle pose is estimated by minimizing both errors of scan-matching and odometry, simultaneously. The localization is achieved even in the environment that the pose cannot be localized uniquely by only the scan-matching method since the information for the estimation is complemented by the odometry. We verify the effectiveness of the proposed method through the off-line localizations using the experimental data. The superiority of the proposed method is shown in a comparison with an extended Kalman filter based method.

There is a demand for autonomous driving control in front-wheel steering vehicles because it is expected to make driving safer and easier and also to reduce the driving workload. In order to perform safe driving with autonomous driving control, it is necessary to consider unexpected disturbances when the vehicle is moving and that tire forces have limitations. We propose autonomous driving control combining Model Predictive Control (MPC) and Sliding Mode Control (SMC). In this paper, we employ MPC in order to consider the maximum tire forces. SMC is employed to deal with unexpected disturbances that the model has not anticipated. Furthermore, we confirmed that path following control is possible by practical inspection using a small front-wheel steering vehicle that is susceptible to unexpected disturbances.

This paper reveals the minimal number of inputs for hovering control from the perspective of nonlinear control system theory. Then, a nonlinear controller is designed to obtain meaningful motion for an underactuated UAV with the consideration of the limitation acknowledged through the model analysis. Throughout this paper, a particular UAV model having three rotors is considered as a controlled object. First, nonlinear state equations are computed with valid assumptions. Next, the model analysis reveals that hovering control is impossible for not only the proposed model but also a generic three-inputs model through the argument of Locally Asymptotically Stabilizability(LAS). Finally, numerical simulation confirms that position control is realizable via Output Zeroing Control along with preferable attitude.

The objective of this paper is to present systematic methods for throwing motion control of underactuated robots. The Pendubot, an underactuated two-link planar robot, is investigated as a dynamic model of the superior limbs for imitation of human throwing motion whose models are fundamentally underactuated in nature. The controller is designed based on input-output linearization and output zeroing control since the Pendubot is not input-state linearizable. The originality of this paper is to intentionally destabilize the zero dynamics, a nonlinear dynamics which remains unobservable from the output when the partially linearized dynamics converges exponentially to zero, to generate dynamic acceleration of the ball. Exact analysis of ordinary differential equations guarantees the explosive instability of the zero dynamics. Numerical simulations confirm the effectiveness of the proposed control strategy.

Novel approach to linearize a multi-input nonlinear system is introduced and applied to control a three-link acrobot. The proposed linearizing method, referred as hierarchical linearization, consists of two ideas; directional decoupling and autonomization. Using these ideas, subsystems and the control inputs for these subsystems are constructed hierarchically. In this paper, the hierarchical linearization is applied to a three-link acrobot and linearizes the system partially. Owing to the proposed method, stable zero dynamics is designed. In numerical simulation, the validity of the controller designed by the proposed method is verified.

In view of the fact that actuators of satellites often break down in long term missions, this paper tackles attitude control problems of satellites by two actuators, especially two reaction wheels, for a failsafe. The contributions of this work, compared with our previous works, are to consider not only kinematics but dynamics of the satellite and to handle arbitrary structure of the communication antenna and two wheels axes. Based on the analysis on equilibrium points, rotation movement and controllability, we divide the control problem into the following three cases. First, in the case that equilibrium points exist, we propose a controller to achieve arbitrary directions of the communication antenna without any rotation. Secondly, if there exists no equilibrium point, we give a sufficient condition and a control law to achieve the desired direction of the antenna, where the satellite remains to rotate around the desired axis. Finally, when neither the first case condition nor the second one is satisfied, it is known that we cannot keep the direction of the antenna in the desired direction, and we can control the direction periodically instead.

Measurement using 3D static cameras can achieve high accuracy localization, but outlier or occlusion should be considered. It is necessary to compensate them to improve accuracy. To address this issue, we introduce Moving Horizon Estimation (MHE) and compare it with Extended Kalman Filter (EKF) to evaluate the estimation accuracy. In this paper, we conduct 3D static camera measurement for a vehicle under challenging conditions in both numerical simulation and experiment. Then, through estimation of position and heading angle of the vehicle, the estimation accuracy is compared to show the effectiveness of the state estimation by MHE even under occlusion of images.

In this paper, we propose a parking control method with on-line path generations and multiple switching motions using path-following control for front steering vehicles. In the proposed method, the parking control is accomplished by following the reference path which is generated on-line adapting to the surrounding environment. Furthermore, we also realize the multiple switching motions to achieve the parking control in difficult situation. The performance of the proposed method is verified through experiments using a 1/10 scale vehicle. In addition, we show that the feasible area of the parking control is extended comparing with the conventional one.

In this paper, we propose a model predictive obstacle avoidance control method for leg/wheel robots with 3 DOF SCARA legs. This robot is able to pass through a narrow space because it can extend and shrink the size between the wheels by changing the configuration of the legs. In this study, obstacle avoidance is achieved by two passage width constraints. Obstacle size and wheel positions are considered in those constraints. A distance between the robot and an obstacle is kept larger than a value specified by those constraints. Comparing with conventional artificial potential methods, the proposed method does not require tuning of obstacle parameters. We utilize two kinds of linear constraints. In the first constraint, we use the internal common tangent between the robot and the obstacle nearest to the robot. In the other constraint, we use the external common tangent to reduce getting stuck into the narrow space between obstacles. The optimal tracking control is realized using those constraints for the model predictive control. The efficacy of this method is verified via the numerical simulations.

In order to increase work efficiency, alleviating burden of operators is important. An autonomous hydraulic excavator is expected to improve it. In this paper, an automatic control of a digging operation for the hydraulic excavator is studied. We propose a method for the trajectory tracking control using model predictive control (MPC) which incorporates servo mechanism. MPC can optimize motion and avoids rapid change of velocity using constraints. However, it is difficult to cope with unknown reaction forces caused by contacting with underground objects. Servo mechanism suppresses the disturbance by the integration of the tracking error. However, the error may be accumulated in the integration. Hence, the trajectory tracking may result in rapid response when the objects are removed. By combining MPC and servo mechanism, we can expect that servo mechanism works against the disturbance and the tracking performance is improved. We show effectiveness of the proposed method through simulations under the presence of the disturbance.

In this paper, we describe a model predictive control (MPC) for small unmanned aerial vehicles (UAVs) like the AR. Drone. By storing the disturbance information, we share it to other UAVs on the same path so that the other UAVs can know the disturbance in advance. Moreover, other UAVs are possible to use the disturbance information in the prediction steps. In addition, the prior disturbance information can be included in the model, and we use the model predictive control to perform the optimization calculation. As the result, the UAVs can prepare to the disturbance before it affects the UAVs, which make them possible to improve the stability of the flight. We verify this approach through the simulation and experiment.

In this paper, we propose the nonlinear controller which stabilizes the attitude of underactuated spacecraft equipped with two reaction wheels in the case of the non-zero initial angular momentum. The attitude of spacecraft is expressed by Rodrigues parameters, and Rodrigues parameters has the property that the attitude converges to specific one when the magnitude of Rodrigues parameters goes to infinity with keeping their ratio to the specified value. To raise the magnitude and keep the ratio, we decoupled two inputs, and design controllers for kinematics model [11][12]. However, in the case of dynamics model, we cannot decouple the inputs for the reason that the decoupling matrix becomes singular. Thus, we use the hierarchical technique which changes geometric nature by designing inputs in step by step. Using the hierarchical technique, the attitude control of underactuated spacecraft can be achieved since the linearization of the system between input and output is possible. As an advantage of our proposed method, the hierarchical technique can linearize the system up to five out of six dimensions whereas the conventional input-output linearization can do up to four dimension. Therefore, we can easily analyze the stability of system including unobservable zero dynamics. Finally, the numerical simulation shows effectiveness of proposed method.

The attitude control of a spacecraft utilizing only two reaction wheels is proposed in [5]. However, the model error and initial angular momentum are not considered in the model base controller, and that may become a problem when it has been applied to actual spacecraft. In this study, considering the case of unknown initial angular momentum and a model error existing in inertia matrix, we propose an adaptive modification of model parameters. The validity of the proposed method is verified by numerical simulations.

In this paper, we cope with the JSAE-SICE benchmark study which demands recovering disadvantage of micro electric vehicles. To enhance the stability of micro cars, we apply the robust path tracking control combining Model Predictive Control and Sliding Mode Control. In this controller, the friction circles are considered using the approximated tire force usage rate to balance the path following performance and the robustness. The enhancement of the vehicle stability using the robust path tracking controller through the numerical simulations is shown. The controller are applied to the all scenarios of the JSAE-SICE benchmark study.

The attitude of a spacecraft is usually controlled by more than 3 reaction wheels. Therefore, as a preparation for accidents of the wheels, it is important to consider the attitude control by 2 wheels. The purpose of this analysis is to find the feasible attitude by two reaction wheels under an initial angular momentum. When the initial angular momentum is not zero, a spacecraft has a drift term and the attitude cannot be controlled freely. Therefore, based on the equilibrium points analysis, this paper proposes the best antenna arrangement in terms of the attitude control by two reaction wheels. Moreover, a design algorithm of a controller that stabilizes a feasible attitude is provided by modifying our previous method.

Omnidirectional vehicles with independent driving and steering wheels are useful since they can move with arbitrary attitude and change the moving direction rapidly. However, excessively rapid steering and unacceptable load may damage mechanisms. To prevent rapid steering and overloads, low-pass filter (LPF) is applied to the steering signals. Since the angular velocity should be limited on account of the motor constraint, even if the time constant is set small, the convergence of the steering is slow, and the error between the target angle and the real steering angle may occur. In this study, we propose the model predictive steering control for independent four-wheel driving and steering vehicles with coaxial steering mechanisms. The proposed method can consider physical constraints of actuators and achieve smooth and quick convergence of the steering to the target angle. We verify the efficiency of the proposed method by experiments using the vehicle which is equipped with an embedded CPU.

In this paper, velocity estimation method for caster odometers is proposed. Caster odometer is a sensor of travel distance and attitude angle using caster mechanism that has two-degree of rotational freedoms. We calculate velocities of a vehicle utilizing the output measured by caster odometers, however, the caster mechanism has nonlinearity component. To deal with the nonlinearity of the caster mechanism, a velocity is estimated by the extended Kalman filter. It is expected to reduce the effect of noise with higher precision than conventional methods (approximated differentiation, extended Luenberger observer). We confirm the accuracy of the proposed estimation by comparing to the conventional method in numerical simulation.

In our previous report on an integrated localization using laser range scanners (LRS) and odometry, a variable covariance depending on the number of data, predicted incident angle, and predicted distance of LRS is introduced to EKF to deal with their effects on the localization. In the present paper, this method is applied to a front steering vehicle navigated by the LRS localization and odometry data. Its experimental results show that the proposed method localize the vehicle position and the outlier of LRS data is properly vanished.

In this paper, a nonlinear controller is designed for an underactuated UAV having three inputs to realize “circle motion”, the motion that the attitude of UAV points to the center of circle while the center of mass moves on a circle orbit. The proposed method is redesigning the time-varying output zeroing controller discretely. This method allows output functions to be kept zero even though output functions are changed. In other words, zero dynamics can be controlled by this method. Finally, numerical simulation shows the validity of the proposed nonlinear controller.

In this study, we apply a hierarchical model predictive control to omni-directional mobile vehicle, and improve the tracking performance. We deal with an independent four-wheel driving/steering vehicle (IFWDS) equipped with four coaxial steering mechanisms (CSM). The coaxial steering mechanism is a special one composed of two steering joints on the same axis. In our previous study with respect to IFWDS with ideal steering, we proposed a model predictive tracking control. However, this method did not consider constraints of the coaxial steering mechanism which causes delay of steering. We also proposed a model predictive steering control considering constraints of this mechanism. In this study, we propose a hierarchical system combining above two control methods for IFWDS. An upper controller, which deals with vehicle kinematics, runs a model predictive tracking control, and a lower controller, which considers constraints of coaxial steering mechanism, runs a model predictive steering control which tracks the predicted steering angle optimized an upper controller. We verify the superiority of this method by comparing this method with the previous method.

The measurement using a motion capture camera is fluctuated by white noise and outliers. In addition, markers to be measured are frequently hidden from cameras by occlusion, then the position and heading angle of a vehicle cannot be uniquely determined because of failure to detect sufficient number of markers. Thus, robust estimation method is required which suppresses the influence of the white noise, the outlier and the occlusion. In this study, we introduce Moving Horizon Estimation (MHE) using partial marker information of motion capture system. It optimizes the objective function using both the marker information in the evaluation range and the constraints on the robot dynamics. By virtue of introduction of constraints, even if the cameras fail to measure the actual state of the robot, the estimated value is determined by MHE. It is the difference from our previous research which assumed that sufficient number of markers are available. In this paper, we estimate the position of the vehicle robot by MHE using the information of the measured markers on the robot, even if several markers are hidden. We will prove the effectiveness of the proposed method by comparing MHE with EKF.

Recently, the increase of burden to operators and lack of skilled operators are the issue in the work of the hydraulic excavator. These problems are expected to be improved by autonomous control. In this paper, we present experimental results of hydraulic excavators using model predictive control (MPC) which incorporates servo mechanism. MPC optimizes digging operations by the optimal control input which is calculated by predicting the future states and satisfying the constraints. However, it is difficult for MPC to cope with the reaction force from soil when a hydraulic excavator performs excavation. Servo mechanism suppresses the influence of the constant disturbance using the error integration. However, the bucket tip deviates from a specified shape by the sudden change of the disturbance. We can expect that the tracking performance is improved by combining MPC and servo mechanism. Path-tracking controls of the bucket tip are performed using the optimal control input. We apply the proposed method to the Komatsu- made micro hydraulic excavator PC01 by experiments. We show the effectiveness of the proposed method through the experiment of digging soil by comparing servo mechanism and pure MPC with the proposed method.

Obstacle avoidance is one of the important tasks for mobile robots. In this paper, we study obstacle avoidance control for mobile robots equipped with four legs comprised of three DoF SCARA leg/wheel mechanism, which enables the robot to change its shape adapting to environments. Our previous method achieves obstacle avoidance by model predictive control (MPC) considering obstacle size and lateral wheel positions. However, this method does not ensure existence of joint angles which achieves reference wheel positions calculated by MPC. In this study, we propose a model predictive control considering reachable mobile ranges of wheels positions by combining multiple linear constraints, where each reachable mobile range is approximated as a convex trapezoid. Thus, we achieve to formulate a MPC as a quadratic problem with linear constraints for nonlinear problem of longitudinal and lateral wheel position control. By optimization of MPC, the reference wheel positions are calculated, while each joint angle is determined by inverse kinematics. Considering reachable mobile ranges explicitly, the optimal joint angles are calculated, which enables wheels to reach the reference wheel positions. We verify its advantages by comparing the proposed method with the previous method through numerical simulations.

Localization is an important function for the robots to complete various tasks. For localization, both internal and external sensors are used generally. The odometry is widely used as the method based on the internal sensors, but it suffers from cumulative errors. In the method using the laser range sensor (LRS) which is a kind of external sensor, the estimation accuracy is affected by the number of available measurement data. In our previous study, we applied moving horizon estimation (MHE) to the vehicle localization for integrating the LRS measurement data and the odometry information where the weightings of them are balanced relatively adapting to the number of the available LRS measurement data. In this paper, the effectiveness of the proposed localization method is verified through both numerical simulations and experiments using a 1/10 scale vehicle. The verification is conducted in the situations where the vehicle position cannot be localized uniquely on a certain direction using the LRS measurement data only. We achieve accurate localization even in such a situation by integrating the odometry and LRS based on MHE. We also show the superiority of the method through comparisons with a method using extended Kalman filter (EKF).

This paper considers the problem to transfer the state from one zero dynamics submanifold to another one in finite time, for a time-invariant nonlinear system. The usage of time-varying zero dynamics submanifold is proposed to accomplish the transfer. The feature of this paper is facus on keeping the state on zero dynamics submanifold during the transfer. Main contribution is to develop the condition for doing this for the case that the two zero dynamics submanifolds have the same dimension. The validity of the controller that is designed to satisfy the condition is demonstrated via a numerical simulation of mono-rotor unmanned aerial vehicle (UAV) system.

In this paper, trajectory tracking and balancing control for autonomous bicycles with a balancer are discussed. In the proposed control method, an input-output linearization is applied for trajectory tracking control and a nonlinear stabilizing control is used for the balancing control. Even though control methods are designed independently, it is shown by several numerical simulations and experiments using a detail model and a real electric motor bike that the stability of the bicycles is ensured with the method even when the desired speed is zero and trajectory tracking to desired ones are achieved.