Kazuyuki Kobayashi

Development of laser rangefinder-based SLAM algorithm for mobile robot navigation

This paper describes a new implementation of the SLAM algorithm for a mobile robot operating in an outdoor environment such as the IGVC Navigation Challenge, using relative obstacle observation profile from laser rangefinder. The proposed SLAM is possible for the mobile robot to start in an unknown location in an unknown environment and, using relative observations only, incrementally build a perfect map of the world and to compute simultaneously a bounded estimate of mobile robot location by the extended Kalman filter. To confirm the proposed SLAM method, an electric wheelchair based mobile robot is used for implementation and testing.

Rotation Angle Measurement of High-speed Flying Object

This paper describes how to measure the rotation angle of a high-speed flying object in 3D space. As the flying object, we consider a golf ball just after impact, whose speed ranges from 40 m/s to 80 m/s. As the sensing device, we employed a high-speed one-line CCD camera instead of employing multiple synchronized 2D image sensors. On the equator and on its perpendicular great circle, black markers are painted to identify rotation angle. The line sensor captures the golf ball and markers when the ball passes through the image capturing plane. From the image, we estimate the rotation angle of the ball. The quaternion approach is used to estimate the rotation angle instead of the Euler angle expression to avoid the gimbal lock phenomenon. Simulations and experiments were carried out to show the validity of the method, and reasonable results were obtained

Topological Map Generation based on Delaunay Triangulation for Mobile Robot

This paper describes a new topological map generation method for mobile robots. The proposed method requires only a laser rangefinder and can automatically generate topological maps based on Delaunay triangulation. By applying triangulation, allowable space can be easily and automatically detected. Based on detected allowable space, facilities structures are abstracted as landmark triangles, which, when detected, are analyzed to generate an appropriate allowable path and moving direction. The topological nodes based on landmark triangles can be converted into a local map as a useful tool for indoor mobile robot navigation. To confirm the proposed topological map generation method, an electric wheelchair based mobile robot is used for implementation and testing. Experiments in an indoor environment demonstrate the validity and feasibility of the proposed methods

Clothoidal Curve-based Path Generation for an Autonomous Mobile Robot

A new path generating method for mobile robots which is equipped with an omni-directional camera and a laser rangefinder is presented. The omni-directional camera can be used for environmental recognition, and can capture 360-degree images without a dead angle. The laser rangefinder is capable of detecting range profiles that can accurately identify the distance of obstacles. This information can be used for smooth path generation. Here, in order to achieve smooth path generation, we applied omni-directional image and laser range profile data to generate an appropriate path by using a clothoid. The curvature and its derivative are estimated by a segmented Hough transform, and path points are generated based on the Delaunay triangulation method. The path generated by the clothoid minimizes the effect of lateral force on the passenger. The validity of the proposed method is verified by simulations

Kinematics simulation by using ODE and MATLAB

This paper describes the kinematics of physical model simulation by using the open dynamics engine (ODE) and matrix based simulation by using MATLAB. In order to confirm the simulation results of the physical model, we derive a mathematical model and perform simulation by using both ODE and MATLAB. Independent simulation model-building and simulation by using both ODE and MATLAB can avoid careless errors. We used this approach to build biped robot with 10 degrees of freedom, and confirmed the validity of the physical simulations by both ODE and MATLAB simulations.

Vision-based Two-phase Strategy for Autonomous Lane Navigation

A new practical autonomous lane navigation strategy for mobile robots, that called vision-based two-phase strategy, is proposed. The proposed strategy consists of a two-phase navigation process. In the first phase, the mobile robot acquires a time-series of images and completes the image-processing iterations until it has sufficient confidence of success in generating the appropriate path. During image acquisition and calculation, the vehicle does not move. Subsequently, the mobile robot navigates based on the appropriate path provided by the processing in the first phase. The key strategy in controlling the mobile robot is the separation of offline route scheduling and online navigation control given by the two phases. To confirm the proposed strategy, we implement a real autonomous electric wheelchair that can be guided by captured images. The validity of the proposed strategy is examined by actual experiments

A Multi-variable Detecting Sensor and its Application

Any sensor to detect a certain physical variable is influenced to some degree by other physical variables designated as noise. The objective in conventional sensor design is to minimize such noise. In this study, conversely we welcome sensing devices that are easily influenced by many physical variables and make full use of their multi-sensing characteristic. We consider such devices as multiple-inputs and single-output sensors. The output signal derived from multiple input signals must be dissociated. The input signals resulting from physical realities have inherent characteristics and can be mathematically modeled. Application of a Kalman filter realized by such models can provide estimates of state variables of all input models, and thus the input signals can be dissociated. As an example, a novel sensor based on a microphone is presented. It can detect various variables such as pressure and acceleration in the frequency range of 0.1 Hz to 10 kHz, temperature, and even light emission. We use the sensor to monitor the symptoms of fire, earthquake and break-in by intruders from within a house

A study of omni-directional image based environment recognition for mobile robots

The development of intelligent autonomous mobile robots is an area of active research. To develop intelligent autonomous mobile robots, it is necessary to select appropriate sensing devices and a robust environment recognition algorithm. To achieve robust and stable navigation for mobile robots, we employ an omni-directional camera, which is a suitable image capturing device for mobile robots because it can capture images of the surrounding environment without a dead angle. In order to recognize the surrounding environment from captured images, we newly developed an environment recognition algorithm that is suitable for omni-directional images. From the results of environment recognition, the robot can find an appropriate path for navigation. The features of the new algorithm are as follows: (1) Morphological operations are applied to detect obstacles robustly and accurately; (2) A quad-tree region segmentation based lane detection algorithm is applied; and (3) An improved Hough transform which can detect curved lanes is used to confirm the results of (2). The effectiveness of the new algorithm was confirmed.

Development of intuitive visual navigation interface for electric wheelchair

In this paper, we develop a new intuitive visual oriented control interface for electric wheelchair instead of conventional joystick interface. The proposed visual-oriented control interface consisting of a head mounted display (HMD) with a two-dimensional acceleration sensor to measure angle based on ground gravity. In order to control the direction and/or speed of the wheelchair, head angle is used as an intuitive driving control interface, which is implemented on a newly developed fuzzy-logic based PSoC microcontroller. By applying fuzzy-logic, smooth and intuitive control can be achieved. The HMD is an intuitive information display with 360-degrees vision and accurate range information.

Foundation-Applied Skin Color Prediction Via SVD-Based Neural Network

When selecting foundation at a cosmetics store, the main objective is determining which material color best suits the users skin color. This is not an easy task without actually applying the foundation since the color observed in the case is not necessarily being the color once it is applied to the skin. We propose a foundation-applied skin color prediction method using an SVD-based artificial neural network. Using the trained network, we achieved 95% of the foundation-applied skin color within a 2-color difference in L*a*b* space