鈴木 智

<jats:p>In the past few decades, drones have become lighter, with longer hang times, and exhibit more agile performance. To maximize their capabilities during flights in complex environments, researchers have proposed various model-based perception, planning, and control methods aimed at decomposing the problem into modules and collaboratively accomplishing the task in a sequential manner. However, in practical environments, it is extremely difficult to model both the drones and their environments, with very few existing model-based methods. In this study, we propose a novel model-free reinforcement-learning-based method that can learn the optimal planning and control policy from experienced flight data. During the training phase, the policy considers the complete state of the drones and environmental information as inputs. It then self-optimizes based on a predefined reward function. In practical implementations, the policy takes inputs from onboard and external sensors and outputs optimal control commands to low-level velocity controllers in an end-to-end manner. By capitalizing on this property, the planning and control policy can be improved without the need for an accurate system model and can drive drones to traverse complex environments at high speeds. The policy was trained and tested in a simulator, as well as in real-world flight experiments, demonstrating its practical applicability. The results show that this model-free method can learn to fly effectively and that it holds great potential to handle different tasks and environments.</jats:p>

<jats:p>In recent years, multi-rotor unmanned aerial vehicles (UAV) have been widely applied for various applications; however, they are yet to be as commonly utilized in certain industrial transportation applications. Thus, this work designed and implemented a reference model-based integral sliding mode control (SMC) method applied to the velocity controller of a multi-rotor UAV. The designed controller was compared with an integral SMC scheme, and then the controller and modeling robustness were verified. Finally, the proposed method was applied to an industrial six-rotor UAV. Three experiments involving target-tracking, fixed-point hovering, and robustness verification were executed. A load of approximately 81.5 % of the UAV’s self-weight was used to verify the robustness of the proposed scheme against parameter uncertainty. This work will serve as a meaningful reference for the application of the SMC in practical industrial applications.</jats:p>

<jats:p>As the scale of the power grid continues to expand, the human-based inspection method struggles to meet the needs of efficient grid operation and maintenance. Currently, the existing UAV inspection system in the market generally has short endurance power time, high flight operation requirements, low degree of autonomous flight, low accuracy of intelligent identification, slow generation of inspection reports, and other problems. In view of these shortcomings, this paper designs an intelligent inspection system based on self-developed UAVs, including autonomous planning of inspection paths, sliding film control algorithms, mobile inspection schemes and intelligent fault diagnosis. In the first stage, basic data such as latitude, longitude, altitude, and the length of the cross-arms are obtained from the cloud database of the power grid, while the lateral displacement and vertical displacement during the inspection drone operation are calculated, and the inspection flight path is generated independently according to the inspection type. In the second stage, in order to make the UAV’s flight more stable, the reference-model-based sliding mode control algorithm is introduced to improve the control performance. Meanwhile, during flight, the intelligent UAV uploads the captured photos to the cloud in real time. In the third stage, a mobile inspection program is designed in order to improve the inspection efficiency. The transfer of equipment is realized in the process of UAV inspection. Finally, to improve the detection accuracy, a high-precision object detector is designed based on the YOLOX network model, and the improved model increased the mAP0.5:0.95 metric by 2.22 percentage points compared to the original YOLOX_m for bird’s nest detection. After a large number of flight verifications, the inspection system designed in this paper greatly improves the efficiency of power inspection, shortens the inspection cycle, reduces the investment cost of inspection manpower and material resources, and successfully fuses the object detection algorithm in the field of high-voltage power transmission lines inspection.</jats:p>

To ensure the safety of autonomous Multi-rotor UAVs flying in urban airspace, they should be capable of avoiding collisions with unpredictable dynamic obstacles, such as birds. UAVs must consider both relative position and relative velocity to avoid moving obstacles. Model predictive control (MPC) can consider the multiple collision avoidance constraints in a constrained optimisation framework. This study proposes a chance-constraints based on obstacle velocity (CCOV) method, which can be combined with previous positional chance constraint methods to account for uncertainty in both position and velocity. This effectively prevents collision with high-velocity obstacles, even in a noisy environment. The proposed method has been performed on a numerical simulation built in MATLAB.

The utilization of small unmanned aerial vehicles (SUAVs), commonly known as drones, has increased drastically in various industries in the past decade. Commercial drones face challenges in terms of safety, durability, flight performance, and environmental effects such as the risk of collision and damage. Biomimetics, which is inspired by the sophisticated flying mechanisms in aerial animals, characterized by robustness and intelligence in aerodynamic performance, flight stability, and low environmental impact, may provide feasible solutions and innovativeness to drone design. In this paper, we review the recent advances in biomimetic approaches for drone development. The studies were extracted from several databases and we categorized the challenges by their purposes-namely, flight stability, flight efficiency, collision avoidance, damage mitigation, and grasping during flight. Furthermore, for each category, we summarized the achievements of current biomimetic systems and then identified their limitations. We also discuss future tasks on the research and development associated with biomimetic drones in terms of innovative design, flight control technologies, and biodiversity conservation. This paper can be used to explore new possibilities for developing biomimetic drones in industry and as a reference for necessary policy making.

In city-wide weather prediction, wind gust information can be obtained using unmanned aerial vehicles (UAVs). Although wind sensors are available, an algorithm-based active estimation can be helpful not only as a weightless substitute but also as feedback for robust control. This paper aims to estimate the wind gusts affecting the quadrotors (a type of UAV) as the input disturbances by using a frequency-based nonlinear disturbance observer (NDOB). To obtain highly accurate estimations, frequency is considered as the main design parameter, thereby focusing the estimation on the frequency range of the wind gusts. The NDOB is developed using the Takagi-Sugeno (T-S) fuzzy framework. In this approach, the twelfth-order nonlinear model is approximated into a sixth-order T-S fuzzy model to reduce computational cost. A two-step verification method is presented, which includes MATLAB/Simulink simulations and the experiments performed using a 2.5 kg quadrotor.

Due to the differences between simulations and the real world, the application of reinforcement learning (RL) in drone control encounters problems such as oscillations and instability. This study proposes a control strategy for quadrotor drones using a reference model (RM) based on deep RL. Unlike the conventional studies associated with optimal and adaptive control, this method uses a deep neural network to design a flight controller for quadrotor drones, which can map the drone's states and target values to control commands directly. The method was developed based on a deep deterministic policy gradient (DDPG) algorithm combined with the deep neural network. The RM was further employed for the actor-critic structure to enhance the robustness and dynamic stability. The RM-DDPG-based flight-control strategy was confirmed to be practicable through a two-fold experiment. First, a quadrotor drone model was constructed based on an actual drone, and the offline policy was trained on it. The performance of the policy was evaluated via simulations while confirming the transition of system states and the output of the controller. The proposed strategy can eliminate oscillations and steady error and can achieve robust results for the target value and external interference.

This article describes the artificial intelligence (AI) component of a drone for monitoring and patrolling tasks associated with disaster relief missions in specific restricted disaster scenarios, as specified by the Advanced Robotics Foundation in Japan. The AI component uses deep learning models for environment recognition and object detection. For environment recognition, we use semantic segmentation, or pixel-wise labeling, based on RGB images. Object detection is key for detecting and locating people in need. Since people are relatively small objects from the drone perspective, we use both RGB and thermal images. To train our models, we created a novel multispectral and publicly available data set of people. We used a geo-location method to locate people on the ground. The semantic segmentation models were extensively tested using different feature extractors. We created two dedicated data sets, which we have made publicly available. Compared with the baseline model, the best-performing model could increase the mean intersection over union (IoU) by 1.3%. Furthermore, we compared two types of person detection models. The first one is an ensemble model that combines RGB and thermal information via "late fusion"; the second one is a 4-channel model that combines these two types of information in an "early fusion" manner. The results suggest that the 4-channel model had a 40.6% increase of average precision for stricter IoU values (0.75) compared with the ensemble model and a 5.8% increase in the average precision compared with the thermal model. All models were deployed and tested on the NVIDIA AGX Xavier platform. To the best of our knowledge, this study was the first to use both RGB and thermal data from the perspective of a drone for monitoring tasks.

With the development of science and technology, the traditional industrial structures are constantly being upgraded. As far as drones are concerned, an increasing number of researchers are using reinforcement learning or deep learning to make drones more intelligent. At present, there are many algorithms for object detection. Although many models have a high accuracy of detection, these models have many parameters and high complexity, making them unable to perform real-time detection. Therefore, it is particularly important to design a lightweight object detection algorithm that is able to meet the needs of real-time detection using UAVs. In response to the above problems, this paper establishes a dataset of six animals in grassland from different angles and during different time periods on the basis of the remote sensing images of drones. In addition, on the basis of the Yolov5s network model, a lightweight object detector is designed. First, Squeeze-and-Excitation Networks are introduced to improve the expressiveness of the network model. Secondly, the convolutional layer of branch 2 in the BottleNeckCSP structure is deleted, and 3/4 of its input channels are directly merged with the results of branch 1 processing, which reduces the number of model parameters. Next, in the SPP module of the network model, a 3 x 3 maximum pooling layer is added to improve the receptive field of the model. Finally, the trained model is applied to NVIDIA-TX2 processor for real-time object detection. After testing, the optimized YOLOv5 grassland animal detection model was able to effectively identify six different forms of grassland animal. Compared with the YOLOv3, EfficientDet-D0, YOLOv4 and YOLOv5s network models, the mAP_0.5 value was improved by 0.186, 0.03, 0.007 and 0.011, respectively, and the mAP_0.5:0.95 value was improved by 0.216, 0.066, 0.034 and 0.051, respectively, with an average detection speed of 26 fps. The experimental results show that the grassland animal detection model based on the YOLOv5 network has high detection accuracy, good robustness, and faster calculation speed in different time periods and at different viewing angles.

This paper presents an airflow vector sensor for drones. Drones are expected to play a role in various industrial fields. However, the further improvement of flight stability is a significant issue. In particular, compact drones are more affected by wind during flight. Thus, it is desirable to detect air current directly by an airflow sensor and feedback to the control. In the case of a drone in flight, the sensor should detect wind velocity and direction, particularly in the horizontal direction, for a sudden crosswind. In addition, the sensor must also be small, light, and highly sensitive. Here, we propose a compact spherical airflow sensor for drones. Three highly sensitive microelectromechanical system (MEMS) differential pressure (DP) sensor chips were built in the spherical housing as the sensor elements. The 2D wind direction and velocity can be measured from these sensor elements. The fabricated airflow sensor was attached to a small toy drone. It was demonstrated that the sensor provided an output corresponding to the wind velocity and direction when horizontal wind was applied via a fan while the drone was flying. The experimental results demonstrate that the proposed sensor will be helpful for directly measuring the air current for a drone in flight.

© 2019 Int. J. Mech. Eng. Rob. Res. In order to traverse the Lunar and Martian surface, planetary exploration rovers, which is equipped with cylindrical typed wheels, has been required high traveling performance. However, the cylindrical wheels of conventional rovers are easy to sink or slip on loose soil. Therefore, the rovers cannot move forward or backward in order to escape from the corresponding severe areas. In this study, we focus on an inching worm locomotion method to solve such a problem. The inching worm locomotion is a method that utilizes bearing force, which is generated between the ground and wheel when the wheel shears the ground. A few previous studies have investigated a method to traverse the loose soil. Further, a static sinkage was used to obtain bearing force in previous studies. This study proposes an advanced scheme that uses large sinkage to increase the traction. In order to confirm the effect of bearing force when the wheel sinkage is large, we performed traveling experiments on loose soil. From experimental results, the traveling performance of the robot, which is operated with the proposed scheme, indicated higher than that of the conventional scheme.

Drones have achieved great success in research and development and in many industrial fields in the last five years, and there are enormous expectations for their further utilization. It was predicted that the market size of drone technologies will rapidly expand in the near future and is expected to reach $12 billion by 2021. In such circumstances, drones have become one of the main research topics in the robotics field, and the number of studies has increased. However, there is little research in the literature that has classified the many studies related to drones and introduced research trends. This paper aims to properly classify drone studies in the robotics field and to summarize the research trends for each classification.

This paper presents the formation control of a fleet of three small quadcopters in a motion capture environment. The dynamic model of a single quadcopter is derived for model predictive control (MPC) and then constraints are explained and expressed in an adequate manner to be included in the cost function for the optimization problem to be solved by the C/GMRES method. Two control architectures, centralized and decentralized, were implemented in the ROS framework and tested on the CrazyFlie quadcopter. First performances are assessed for a static reference, the formation regulation problem, then for a dynamic reference, the formation tracking one. Finally, computational cost of the MPC controllers is evaluated.

In this study, a novel robust navigation system for a drone in global positioning system (GPS) and GPS-denied environments is proposed. In general, the drone uses position and velocity information from GPS for guidance and control. However, GPS cannot be used in several environments; for example, GPS exhibits huge errors near buildings and trees, indoor environments. In such GPS-denied environments, a Laser Imaging Detection and Ranging (LIDAR) sensor-based navigation system has generally been used. However, the LIDAR sensor also has a weakness, and it cannot be used in an open outdoor environment where GPS can be used. Therefore, it is advantageous to develop an integrated navigation system that operates seamlessly in both GPS and GPS-denied environments. In this study, an integrated navigation system for the drone using GPS and LIDAR was developed. The design of the navigation system is based on the extended Kalman filter, and the effectiveness of the developed system is verified by numerical simulation and experiment.

In this paper, collision-free guidance control of multiple small unmanned aerial vehicles (SUAVs) is designed. Collision avoidance of the SUAVs should be considered in the control system design for safe operation. Therefore, a guidance control system using a distributed model predictive control (DMPC) is proposed to realize the collision avoidance. A constraint for the relative position vector between the each UAV is considered in the design for efficient avoidance. Small multi-rotor helicopter is considered as controlled object, and the guidance control system is designed by using the translational model of the helicopter. DMPC is designed with three constraints, an input constraint, a state constraint, and a relative position vector constraint. An input constraint and a state constraint realize collision avoidance in input within the constant limits. If the moving path of the one helicopter is significantly affected by the moving path of other helicopter, the relative position vector constraint makes the helicopters exchange their relative position each other. By using these constraints, smooth collision avoidance is realized. The numerical simulation and flight experiment is conducted to verify the effectiveness of designed control system.

© 2014, Iizuka et al.; licensee Springer. The rovers, which some researchers and agencies are developing, have many functions to sense a lot of information from peculiar environments like the lunar surface for localization, path planning and so on. On rough terrain, without artificial ground maintenance, like the lunar surface, the rovers avoid obstacles by using sensors, which they have. However, if the rovers traverse loose soil, like the planetary surface, there exists the sinking behavior. This sinkage is caused by the weight of the rover. At present the rover sensors are unable to sense it. Actually, Mars Exploration Rover (MER, NASA/JPL), during its exploration on Mars, was disabled because of this. If MER could have known the subsidence of each wheel in time, MER may have avoided this catastrophic condition. We therefore propose to have the grouser mechanism, which can detect sinkage of the wheels of rovers, installed on the wheels of rovers. The grouser mechanism is needed to traverse loose soil like lunar surface and mars surface. When the wheel rotates on loose soil, the resistance force from the loose soil is given to the grouser. However, initially, for the wheel contacts on the surface of loose soil, the resistance from the loose soil is small. We will develop the grouser mechanism with a function that can detect the whole range of the effective resistance force from loose soil including this small resistance force. This study moreover carries out experiments with various loads and verifies the effectiveness of the proposed mechanism, using pictures, of contact with the loose soil surface.

In this study, our aim is to realize collision-free guidance control for a small unmanned helicopter. The simultaneous flight of multiple small unmanned helicopters has recently attracted considerable attention for practical operation because of the high efficiency and fault tolerance capability. Collision avoidance should be considered in the guidance system of small helicopters to realize simultaneous flight. The authors adopted nonlinear model predictive control (NMPC) to design a collision-free guidance control system for small unmanned helicopters; collision avoidance was regarded as a state constraint. A hierarchical control structure consisting of an attitude control system and guidance control system was adopted to simplify the overall control system. The authors propose a simple nonlinear translational model of the helicopter to reduce the computational cost of NMPC. The effectiveness of the proposed collision-free guidance control system was verified through both numerical simulation and a flight experiment.

In this study, we aim at realizing autonomous simultaneous flight of multiple small helicopters. In such situation, collision avoidance of the helicopters should be considered in guidance control system to improve safety and reliability of the flight system. In this paper, we construct a collision-free guidance control system for multiple small helicopters. The collision avoidance problem is regarded as a control problem with state restrictions, and the theory of Nonlinear Model Predictive Control (NMPC) is applied to the guidance control system. A simple nonlinear guidance model is used for design of NMPC to reduce the computational cost. A novel position constraint is proposed for optimizing the avoidance trajectory of each helicopter. The effectiveness of the designed control system and proposed constraints are verified by numerical simulation and flight experiment.

In this study, we design an autonomous navigation, guidance and control system for a small electric helicopter. Only small, light-weight, and inaccurate sensors can be used for the control of small helicopters because of the payload limitation. To overcome the problem of inaccurate sensors, a composite navigation system is designed. The designed navigation system enables us to precisely obtain the position and velocity of the helicopter. A guidance and control system is designed for stabilizing the helicopter at an arbitrary point in three-dimensional space. In particular, a novel and simple guidance system is designed using the combination of optimal control theory and quaternion kinematics. The designs of the study are validated experimentally, and the experimental results verify the efficiency of our navigation, guidance and control system for a small electric helicopter.

An attitude sensor mounted on an unmanned system requires to have high accuracy in attitude estimation under a dynamic acceleration environment for autonomous attitude control. To estimate the attitude, an attitude sensor needs to detect the direction of gravity from measured accelerations, which include dynamic accelerations. However, accelerometers are usually sensitive to both gravity and dynamic accelerations. In this paper, two attitude estimation algorithms with a quaternion based on the extended Kalman filter are described and compared. One of the proposed algorithms uses a pre-filter. In this approach, measured accelerations are filtered by the pre-filter and classified into gravity and dynamic accelerations. Using only the measured gravitational accelerations, an EKF can estimate the attitude with high accuracy. In the other approach, the EKF includes a dynamical model of acceleration disturbance. It can estimate both the attitude and the acceleration disturbance. The comparative experiments with hand motion, moving vehicles, and UAVs, were carried out to evaluate these algorithms. In the result of pre-filter method, estimation errors were accumulated in long-term acceleration disturbance. Meanwhile, the EKF algorithm with a dynamical model of acceleration disturbance provided accurate results in each experiment. As a result, we concluded the latter method is well suited to a dynamic acceleration environment for unmanned systems.

In this research, we design the navigation and fully autonomous control system for small electric helicopter. Small and low-accuracy sensors (GPS which cannot output vertical velocity, low-accuracy accelerometers, and barometer) are used for navigation and control. Firstly, INS/GPS/Barometer navigation system is designed to obtain high-accuracy position and velocity of the helicopter. Secondly, 3-dimensional guidance control system is designed by using optimal control theory and quaternion kinematics. Lastly, whole systems designed in this research are validated by the flight experiments. © 2011 IEEE.

In this study, a nonlinear attitude controller for a small unmanned electric helicopter is designed by using quaternion feedback provided by the backstepping control method. First, a quaternion-based multi-input multi-output(MIMO) nonlinear attitude model of a small helicopter is derived. This nonlinear model consists of three parts, namely, the time derivatives of the quaternion, the Euler equation of rotation, and the flapping dynamics of the main rotor and the stabilizer bar. Next, a nonlinear MIMO controller that calculates the desired torque input by the backstepping control method is designed. The control law is modified to exclude a online computation of time derivatives of sensor output. The controller achieves the asymptotic stability of the origin of the attitude error system, and guarantees that the helicopter could track arbitrary desired attitude. Finally, a simulation and experiment are performed, and the results of this simulation and experiment show the effectiveness of the suggested nonlinear MIMO controller.

In this study, a nonlinear adaptive control system for a single-rotor type small helicopter is designed. First, a small autopilot device, which can be applied any single-rotor type helicopters, is developed based on FPGA (Field Programmable Gate Array) board. In the case of small helicopter, the payload limitation is so tight. Therefore, we can only use small and light weight sensors and computer for the control. All the sensors we use for the autopilot device are light enough to be mounted on any small helicopters. the total weight of the autopilot device with box is about 300 g. Next, a nonlinear adaptive attitude controller for small helicopter is designed by using adaptive backstepping method. It is guaranteed to make the attitude of the helicopter follows arbitrary time varying reference attitude, even when the helicopter model has the parameter uncertainties, such as inertia moment or aerodynamic uncertainties. Finally, the effectiveness of the autopilot device and the adaptive attitude controller are verified by simulation with parameter uncertainties, and the flight experiment using heterogeneous small helicopters.

The attitude sensor mounted on unmanned systems is required high accuracy attitude under the dynamic acceleration environment for the autonomous attitude control. In order to estimate attitude by using a quaternion based on extended Kalman filter (EKF) , the attitude sensor detects a direction of the gravity by using a tri-axis accelerometer. However, accelerometers are usually sensitive to both the gravity and dynamic accelerations. In this paper, the dynamic accelerations are dealt with explicitly, and two types of algorithms are proposed. In the first approach, measured accelerations are filtered, and divided into the gravity and dynamic accelerations. Using only the gravity, EKF estimates high accuracy of attitude. This algorithm can be realized with relatively small amount of computation, and have a good effect under the short dynamic acceleration environment. The second approach demonstrates adequate performance with EKF which includes the dynamical model of accelerations. This paper shows a critical comparison of attitude estimation algorithms under the dynamic acceleration environment for the autonomous control of unmanned systems.

In this paper, we design a quaternion-based attitude controller for small electric helicopter. Firstly, we derive a MIMO(Multi-Input Multi-Output) nonlinear attitude model of a small helicopter. In this model, we adopt the quaternion for attitude expression. This model consist of three parts, the dynamics of quaternion, the Euler's equation, and the dynamics of main rotor and stabilizer. Secondly, we design a MIMO controller by using backstepping control method which is one of nonlinear control method. The asymptotic stability of the origin of error system is achieved by this controller. Simulation and experiment results show that the nonlinear controller is more effective than linear controller which was used in previous study.

In this paper, we describe a model based formation flight control of multiple small-scale unmanned helicopters. We design the autonomous formation flight control system as leader-following configuration. In order to achieve good control performance under the system constraint, "model predictive control" is used for the translational position control of follower helicopters. Position constraints such as moving range and collision avoidance problem are considered in the real time optimal control calculation. To achieve a robustness against disturbance, a minimal order disturbance observer is used to estimate the unobsevable state variables and disturbance. The simulation results are presented to show the feasibility of the control strategy. The designed controller is implemented in a laptop on the ground station by C++ without any other optimizing software. The formation flight control experiment is carried out using two helicopters. The experimental result shows the accurate control performance. The position constraint capability is confirmed through the experiment with the single helicopter. Finally, the robustness against wind is verified by the windy condition experiment.

In this paper, we describe a trajectory following control for an unmanned small-scale helicopter that weighs about 9kg. In order to achieve desirable trajectory following performance, we use model-following-type optimal control methods which are extensions of our previous work. In designing the translational guidance controller, we apply a model-following-type model predictive control method. It is important to consider the trade-off between trajectory following performance and attitude angle response in the design. These elements are reflected in the performance index for the design of the model predictive controller. In the altitude controller design, we apply a model-following-type LQI method. Experimental results are presented to show the effectiveness of our control strategy. Accurate control performance is achieved in circular and sinusoidal trajectory following flight experiments under windy conditions.

In this paper, we propose the attitude estimation algorithm under the dynamic acceleration environment. Generally, an attitude sensor has biaxial or triaxial accelerometer in order to measure the direction of gravity. In this configration, the attitude sensor has serious error under the dynamic acceleration environment, because of the measurement error of gravity that is caused by the dynamic acceleration. When we put the attitude sensor on a movable body like an UAV (Unmanned Aerial Vehicle), this kind of error is fatal for the sensor. So, we apply the extended kalman filter algorithm to reduce the estimation error. Firstly, we derive the process model for the kalman filter which is based on quaternion kinematics. Secondly, we design the extended kalman filter by using the process model. Lastly, we show the simulation and experiment result of the estimation algorithm.

In this paper, we propose the autonomous attitude control of Quad Tilt Wing(QTW)-UAV. QTW-UAV can realize the vertical takeoff and landing, and hovering flight that seems to be the helicopter, and the high cruising speed that seems to be the fixed-wing aircraft by changing an angle of a rotors and wings by using tilt mechanism. Firstly, we make the attitude model of QTW-UAV by using identification method. Secondly, we designed attitude control system with kalman filter based LQI control method. And lastly, we show the experiment result. The experiment result shows model based control design are very useful for autonomous control for QTW-UAV.

In this paper, we propose the autonomous hovering control of hobby class small scale unmanned helicopter by using model based control method. Firstly, we make the model of small-scale unmanned helicopter analytically. In many researches, modeling of small-scale unmanned helicopter has been carried out with the system identification method, because some model parameters have never been derived anallytically. But, the model with system identification method is not for general purpose. For this reason, we use the white box modeling method which is anallytical modeling. Secondly, we designed control system with kalman filter based LQI control method. Control system consists of the attitude controller and the velocity controller and position controller. And we connect these three controllers to series. Finally, the hovering control experiment is carried out, and experiment result shows that analytic modeling method and model based control design are very useful for autonomous control for small scale unmanned helicopter.

In this paper, we design a trajectory following controller by using model following based sliding mode control for small-scale unmanned helicopter. Firstly, we derive reference model that shows an ideal response. This reference model follows continuous trajectory like a circular trajectory without delay. Secondly, we design the model following based sliding mode controller that makes the response of real system follow the response of reference model. Lastly, we show the validity of the controller by the simulation and experiment.