Sayaka Kanata

Extraction of Similarities and Differences in Human Behavior using Singular Value Decomposition

Abstract Human behavior has a structure that consists of similarities and differences. Similarities are motion patterns, which are observed universally and are independent of individuals. Differences are particular characters found in individual motions and can represent identities. We proposed a method for extracting similarities and differences in human behavior by using singular value decomposition. Moreover, we described how to analyze the extracted characteristic motions. We applied our methods to several types of human behavior. The data obtained by applying our method proved that similarities and differences in human behavior can be extracted. We also analyzed the extracted characteristic motions in detail.

Measurement model of barometer in ground effect of unmanned helicopter and its application to estimate terrain clearance

To improve reliability of autonomous unmanned helicopter, GPS-INS hybrid navigation systems using Kalman Filter are utilized. Barometer can be integrated into the hybrid navigation system, but the problem is that there is no sufficient model which can describe the relation between the altitude and the air pressure in ground effect. The purpose of this paper is to develop a measurement model of a barometer on an unmanned helicopter which can be used in the hybrid navigation system for autonomous unmanned helicopters. In particular, we propose that the model equation of atmospheric pressure under ground effect can be applied to estimate the terrain clearance of the helicopter. Results of flight experiments show that the proposed method is effective when the helicopter is close to the surface.

GPS-INS-BARO hybrid navigation system taking into account ground effect for autonomous unmanned helicopter

Autonomous unmanned helicopters are useful in disaster response activities, but their reliability needs improvement to broaden its application area. GPS-INS hybrid navigation systems using Kalman Filter have been used to improve the reliability of autonomous unmanned helicopters, but they are not robust against the absence of GPS measurements. The vertical accuracy of GPS measurement is less than horizontal ones. The vertical channel of INS is unstable. Adding the altitude sensor to GPS-INS hybrid navigation system is expected to increase the accuracy and improve reliability. Barometric altimeter is not so accurate, but it is widely used because of its convenience. For unmanned helicopters, though, it is known that the barometer altitude deviate downward from the actual altitude, where the helicopter is close to the ground. This phenomena is due to the ground effect. This paper describes the method to build GPS-INS-BARO hybrid navigation system, which takes into account the ground effect for the autonomous unmanned helicopter. We adopted a centralized approach to integrate different sensors with asynchronous measurement updates. The parameter sets for successful estimation were shown using numerical simulation results.

Estimation of magnetic declination angle using reduced QUEST for an unmanned aerial vehicle

Measurement of the heading angle which is free from magnetic declination using several GPS antennas is widely applied, but it is not suitable for small unmanned aerial vehicles because of their limit of the payload and body length. In this paper, we propose a method to estimate the magnetic declination angle using an inertial measurement unit(IMU) and only one GPS antenna. The method consists of 1) estimation of the acceleration of the motion from GPS measurements and 2) estimation of the magnetic declination angle using the estimated accelerations and IMU measurements. We show that the estimation of the magneticdeclination angle results in reduced QUESTflO], that is an eigenvalue problem of 2 × 2 real matrix. Flight experimental results using an unmanned helicopter demonstrate the effectiveness of the proposed methods.

Sensitivity analysis and influence discussion of estimation errors in rotation parameters in localization of rovers on small planetary bodies with single source of radio waves

Direct investigations are required for future missions to small planetary bodies, in addition to obtain terrain maps and collecting samples from the surface. Rovers, mobile robots on surface of a planetary body, are effective in investigating into several specific points. We proposed a method of localization for navigating rovers to specific features. The proposed method measures two-way range between the rover and the mother spacecraft orbiting around the investigating planetary body using radio waves. The proposed method can be applied with reasonable accuracy on large and small planetary bodies and it can particularly be used with sub-hundred-meter-sized bodies, which conventional methods cannot be applied to. Sensitivity analysis of the proposed method is summarized in this paper, according with estimation errors in rotation parameters of the planetary body. A method of estimation is described here, which takes into account errors of rotational parameters.

REAL-TIME LOCALIZATION OF A ROVER ON A SMALL PLANETARY BODY BASED ON RADIO WAVES

Investigation on small planetary bodies with diameter less than 1 km has been getting a lot more attention recently. Rovers are promising way for surface investigation of small planetary bodies. A radio-wave-based method of localization for such rovers has been proposed, which uses twoway range between the rover and a mother spacecraft. However, it assumed that the rover is stationary on the surface during the localization. In this research, the rover is supposed to move during the localization. Rovers on small planetary bodies move by hopping. It hops, repeatedly bounds on rough terrain, may rotate on the surface and become stationary at some time. This paper describes a method to localize the hopping rovers in real-time. In the proposed method, multiple motions for rovers are modeled at the time-update steps in the Kalman filter. Numerical simulations are summarized assuming a rover on a planetary body with the size of asteroid Itokawa, which suppose that the proposed method provides accuracy of meter-order localization.

Attitude representation for precise 3D terrain mapping with autonomous unmanned helicopter

Autonomous unmanned helicopters are useful in various disaster prevention activities. 3D terrain mapping system is one of the most promising applications of autonomous unmanned helicopters. To improve the accuracy of the terrain mapping, it is necessary to improve the estimation accuracy of the attitude. In this paper, we focused on minimal representations of the attitude, which do not require unprofitable procedures in filtering, such as normalization. We defined generalized Rodrigues parameters, which are generalizations of Rodrigues and modified Rodrigues parameters. Generalized Rodrigues parameters were compared with Euler angles, which are the most well-known minimal representation of the attitude. Through numerical simulation results, properties of attitude estimation systems using those attitude representations are discussed in this paper. Finally, we concluded that generalized Rodrigues parameters are suitable for 3D terrain mapping.

Changes in coordinative structure through human-environmental interaction and its role in estimation of dynamics

Humans are able to acquire information on dynamics that cannot be directly obtained through observation of kinematics. Since the coordinative structure of the human behavior is self-organizing in interactions with the environment, a change in the environment induces a change in the coordination structure. It is posited that humans identify dynamic interactions between the body of the actor and the external environment through the interpolation of coordinated structures for observed behaviors. Coordinative structures consist of invariant part and covariant part. Invariant part is a set of motion change points which are found even if the environment or dynamic interaction has changed. Covariant part is a set of motion change points caused by the environmental changes. In this paper, we examine more closely the changes in coordinative structure induced by changes in dynamic interactions with the environment, and describe a method to extract invariant and covariant part of coordinative structures. Moreover The importance of the change in the coordinated structure in recognition of the dynamic interaction by observers is also described. The results of dynamics recognition experiments revealed that information necessary for the estimation of relative weight is contained in the covariant part of the coordinated structures but not in their invariant regions, and furthermore that timing is also an essential for estimation.

Accurate Localization for Space Rovers and its Computational Reduction

To localize a rover on small planetary body, a method using round-trip propagation delay of radio waves is most promising. In order to improve the estimation accuracy, it is necessary to estimate rotational motion of the small planetary body. A method of localization has been expanded to estimate also the rotational parameters of the small planetary body. The expanded problem includes complex nonlinear and dynamical issues that it cannot be solved analytically. So, the estimation problem has been formulated as an optimization problem to minimize the loss function defined based on the estimation errors derived in comparison with actual measurement data. A solution for the optimization problem has been proposed, which uses Powells conjugate direction method for local searches. Although this solution does not require calculation of derivatives, it requires a large amount of computation since several forward calculations of the state are required for each minimum search. In this paper, a method to select measurement data is described, which provides as accurate estimation as the original results with reducing the computational amount. The main idea of selection is to conserve the sensitivity of the measurement data. The proposed method to select data is compared with the results using the decimated data of equal interval. Simulation results and experimental results are shown to evaluate computational reduction and estimation accuracy.