Chen-Chien James Hsu

A practical nighttime vehicle distance alarm system

This paper presents a practical nighttime vehicle distance measuring method based on a single CCD image. The method combines an image-based distance measuring system. To solve the nighttime feature extraction problem, the proposed method uses two taillights as the feature. Based on the proportionality of similar triangles, distance between a CCD camera and the taillights of the vehicle ahead can be measured. The method focuses on detecting the taillights and differentiating the targeted vehicle from others on the basis of partial image analysis instead of whole image processing. The system is both fast and simple. The accuracy of the proposed method is demonstrated in this paper through experiences.

Run-time efficient observer-based fuzzy-neural controller for nonaffine multivariable systems with dynamical uncertainties

Abstract In this paper, a novel hierarchical structure with run-time efficiency is developed to solve the rule explosion problem of fuzzy-neural network control for a class of uncertain nonaffine multivariable systems. The parameters of the hybrid adaptive controller are on-line tuned by the derived update laws under the constraint that only system outputs are available for measurement. Compared with the previous approaches, the proposed design process is more flexible and requires less computation time. According to the stability analysis, the overall control scheme guarantees that the closed-loop systems can obtain successful system control, effective state observer, and desired tracking performance. Finally, illustrative examples are provided to show the effectiveness of the proposed approach.

Depth map building based on pixel number variation and SURF

This paper presents a method for depth measurement based on Speeded Up Robust Features (SURF) and pixel number variation of CCD Images. A single camera is used to capture two images in different photographing distances, where features in the images are extracted and matched by SURF. To remove false matching points, an Identifying point correspondences by Correspondence Function (ICF) method is adopted to automatically select suitable reference points required for the pixel number variation method. Based on the displacement of the camera at two photographing distances, difference in pixel count between feature points of the objects in the images can be used to determine the photographing distance of the target objects for constructing the depth map by linear interpolation.

Depth measurement based on pixel number variation and Speeded Up Robust Features

This paper presents a method for depth measurement based on Speeded Up Robust Features (SURF) and pixel number variation of CCD Images. A single camera is used to capture two images in different photographing distances, where features in the images are extracted and matched by SURF. To remove false matching points, an Identifying point correspondences by Correspondence Function (ICF) method is adopted to automatically select suitable reference points required for the pixel number variation method. Based on the displacement of the camera at two photographing distances, difference in pixel count between feature points of the objects in the images can be used to determine the photographing distance of the target objects for constructing the depth map.

Enhanced Monte Carlo localization incorporating a mechanism for preventing premature convergence

In this paper, we propose an enhanced Monte Carlo localization (EMCL) algorithm for mobile robots, which deals with the premature convergence problem in global localization as well as the estimation error existing in pose tracking. By incorporating a mechanism for preventing premature convergence (MPPC), which uses a “reference relative vector” to modify the weight of each sample, exploration of a highly symmetrical environment can be improved. As a consequence, the proposed method has the ability to converge particles toward the global optimum, resulting in successful global localization. Furthermore, by applying the unscented Kalman Filter (UKF) to the prediction state and the previous state of particles in Monte Carlo Localization (MCL), an EMCL can be established for pose tracking, where the prediction state is modified by the Kalman gain derived from the modified prior error covariance. Hence, a better approximation that reduces the discrepancy between the state of the robot and the estimation can be obtained. Simulations and practical experiments confirmed that the proposed approach can improve the localization performance in both global localization and pose tracking.

Simultaneous Localization and Mapping with a Dynamic Switching Mechanism (SLAM-DSM)

In this paper, we propose a simultaneous localization and mapping (SLAM) algorithm incorporating a dynamic switching mechanism to switch between FastSLAM 1.0 and 2.0, based on a threshold of effective sample size (ESS). By taking advantages of FastSLAM 1.0 and 2.0 through the proposed dynamic switching mechanism, execution efficiency is significantly improved while maintaining an acceptable accuracy of estimations. To show the effectiveness of our proposed approach in comparison to FastSLAM 1.0 and 2.0, several simulations are demonstrated in this paper.

Robot Intelligence Technology and Applications 4 - Results from the 4th International Conference on Robot Intelligence Technology and Applications

Computationally efficient SLAM (CESLAM) has been proposed to solve simultaneous localization and mapping problem in real-time design. CESLAM first uses the landmark measurement with the maximum likelihood to update the particle states and then update their associated landmarks later. This improves the accuracy of localization and mapping by avoiding unnecessary comparisons. This paper describes a modified version of CESLAM called rapidly operations SLAM (ROSLAM) which improves the runtime even further. We present an empirical evaluation of ROSLAM in a simulated environment which shows that it speeds up previous well known algorithms by 100 %.

New solutions to the premature convergence problem in Monte Carlo localization

In this paper, a new solution towards the premature convergence problem in Monte Carlo Localization for global localization under highly symmetrical environments is proposed. The algorithm employs a “standard direction” to allow particles to move so as to rearrange weights, providing better exploration as a result. Therefore, there are higher opportunities for particles to converge to the real robot pose and prevent premature convergence accordingly. Experiments have verified the proposed algorithm to be reliable and robust by offering notable improvements in the global localization performance.

Object tracking based on hardware/software co-design of particle filter and particle swarm optimization

This paper presents a hardware/software co-design method for a hybrid object tracking algorithm incorporating particle filter (PF) and particle swarm optimization (PSO) based on System On Program Chip (SOPC) technique. Considering both the execution speed and design flexibility, we use a embedded processor to calculate weight for each particle and a hybrid accelerator implemented by hardware to update particles. As a result, execution efficiency of the proposed hardware/software co-design method is significantly improved while maintaining design flexibility for various embedded applications. As soon as prototype testing for a specific problem is completed by using the software weight assignment, full hardware implementation of the weight calculating module can be used to speed up the execution speed.