Browsing by Author "Song, Kai-Tai"

Now showing 1 - 20 of 90

3D Vision for Object Grasp and Obstacle Avoidance of a Collaborative Robot
(2019-01-01)
Song, Kai-Tai
;
Chang, Yu-Hsien
;
Chen, Jen-Hao
;
電控工程研究所
;
Institute of Electrical and Control Engineering
This paper presents a design and experimental study of 3D robotic vision for bin picking and obstacle avoidance. Through the 3D vision algorithm, the robotic picking system is able to analyze the imagery of cluttered objects, classify the objects and estimate the pose of identified objects for grasping. In order to facilitate the robot to work with a human nearby, obstacle avoidance during task execution is developed based on 3D vision. In this design, a RealSense SR300 RGB-D camera is utilized to acquire RGB images and depth images of clustered workpieces. A deep neural network (DNN) approach to object recognition is designed and combined with point cloud segmentation to enhance 3D object-pose estimation for grasping. The robot avoids obstacles to assure safe operation during execution of the bin picking task. Practical experiments using a Techman TM5 6-DOF robot arm show that the proposed method effectively accomplishes obstacle avoidance in pick-and-place operations.
A Client-Server Architecture for Object Volume Measurement on a Conveyor Belt
(2019-01-01)
Song, Kai-Tai
;
Ou, Song-Qing
;
Yung-Ping-Tien
;
電控工程研究所
;
Institute of Electrical and Control Engineering
This paper presents an object volume measurement system on a conveyor belt. A client-server architecture is proposed to obtain a flexible and efficient measurement system for object size on the conveyor belt. The client works to integrate the sensor data and convert them into three dimensional point cloud. The server is responsible for calculating the object size. A laser scanner scans the object on the conveyor belt that moves with a constant speed. We applied the trigonometric function method to integrate two-dimensional distance data form the laser scanner and encoder data of the conveyor belt to obtain three-dimensional point cloud data. A progressive morphological filter was adopted to segment the background point cloud and the object point cloud. Then a convex hull algorithm was applied to find the set of the smallest convex polygon that contains all the object point cloud. Further, we can find a rectangle which has the smallest area to determine the object size. The experimental results show that the shape of object can be square, cylindrical or irregular. When the conveyor belt speed is set at 20 cm/s, the percentage error of the object size measurement is around 5% with the laser scanner installed at 50 cm above the conveyor belt and that within 7% with the laser scanner installed at 100cm.
A Compliance Control Design for Safe Motion of a Robotic Manipulator
(2011-01-01)
Song, Kai-Tai
;
Hsu, Chieh-Hsun
;
電機工程學系
;
Department of Electrical and Computer Engineering
For domestic and service robots, the robotic manipulator may cause danger or damage to surrounding objects and people if the manipulator does not understand the details about the environment. To cope with this problem, we investigate compliance motion controller design in order to develop a dependable and safe robot. This paper presents a novel safe motion strategy together with its experimental study for a robot arm. Taking into account a manipulator in a home environment, we divide manipulator behaviors into standby mode and motion mode. For standby mode, a compliant motion design has been developed for force guide-through motion. In motion mode, the proposed safety motion strategy ensures safe execution of tasks. Experimental results of a 4-DOF manipulator constructed in the lab validate the developed safe motion strategy.
A Design for Integrated Face and Facial Expression Recognition
(2011)
Song, Kai-Tai
;
Chen, Yi-Wen
;
電機工程學系
;
Department of Electrical and Computer Engineering
An integrated face and facial expression recognition system has been designed and tested for robotic applications. Facial images from a web camera are first acquired for facial shape and texture model generation using active appearance model (AAM). Modified Lucas-Kanade image alignment algorithm was adopted to find facial features as well as the texture model of AAM to construct facial texture parameters. These parameters are used to train back propagation neural networks(BPNN) for face and facial expression recognition. A novel design is proposed for an integrated facial expression recognition system. In the first stage, face recognition is performed to find user's identity; then the facial-expression database of the recognized user is employed to recognize his/her facial expressions. Experimental result based on BU-3DFE database show that a face recognition rate of 98.3% is achieved. The facial expression recognition rate of the proposed integrated method (using personal facial expression classifiers) is 83.8%, an improvement compared with 69.6% of using conventional classifiers.
A fast learning algorithm for robotic emotion recognition
(2007)
Hong, Jung-Wei
;
Han, Meng-Ju
;
Song, Kai-Tai
;
Chang, Fuh-Yu
;
電控工程研究所
The capability of robotic emotion recognition is an important factor for human-robot interaction. In order to facilitate a robot to function in daily live environments, a emotion recognition system needs to accommodate itself to various persons. In this paper, an emotion recognition system that can adapt to new facial data is proposed. The main idea of the proposed learning algorithm is to adjust parameters of SVM hyperplane for learning emotional expressions of a new face. After mapping the input space to Gaussian-kernel space, support vector pursuit learning (SVPL) is applied to retrain the hyperplane in the new feature space. To expedite the retraining procedure, only samples classified incorrectly in previous iteration are combined with critical historical sets to restrain a new SVM classifier. After adjusting hyperplane parameters, the new classifier will recognize previous erroneous facial data. Experimental results show that the proposed system recognize new facial data with high correction rates after fast retraining the hyperplane. Moreover, the proposed method also keeps satisfactory recognition rate of old facial samples.
A Fuel Cell Power Supply System Based on Charging Scheduling of Multiple Lithium Batteries
(2018-01-01)
Song, Kai-Tai
;
Song, Shao-Huan
;
Wang, Sz-Sheng
;
電控工程研究所
;
Institute of Electrical and Control Engineering
In recent years, various service-oriented mobile robots have become more and more popular. These mobile robots need electrical power for continuous work for some duration using some kind of battery and hence a battery charging system is required. In order to increase the durability of robot operation, it is desirable to provide a power generation as well as a power supply system onboard the robot. This paper presents a design and implementation of a power supply system combining a fuel cell and lithium batteries to increase durability. It will be a new option to provide durable and environment-friendly energy for mobile robot. In this paper, we propose a charging logic to manage fuel-cell charging of multiple lithium polymer batteries, so that the power supply system can switch among lithium polymer batteries according to the load requirement. Experimental results of the developed power supply system on a laboratory mobile robot verified the effectiveness of the proposed design.
A New Approach to Map Joining for Depth-Augmented Visual SLAM
(2013)
Liu, Chien-Hung
;
Song, Kai-Tai
;
電控工程研究所
;
Institute of Electrical and Control Engineering
In this paper, a novel scheme is proposed to improve real-time performance of simultaneous localization and mapping (SLAM) of a mobile robot based on depth-augmented visual features. In this design, the robot has two stages in navigation applications, namely the map building stage and the map usage stage. In the map-building stage, a local map is built to join into the global map. For the map-usage stage, instead of using the global map, the local maps facilitate real-time path tracking control of the robot. Using of local maps has the merit of reducing the computational complexity of EKF-SLAM. In the map joining procedure, deviations of adjacent local maps are corrected based on local features. Loop closure detection is used to determine whether the local map building is completed. A Kinect sensor is adopted to realize the proposed method. Navigation experiments on a wheeled mobile robot show that the motion error of robot localization is within 0.1m for a travel over 83m.
A New Compliant Motion Control Design of a Walking-Help Robot Based on Motor Current and Speed Measurement
(2009)
Song, Kai-Tai
;
Lin, Chen-Yang
;
電子工程學系及電子研究所
;
Department of Electronics Engineering and Institute of Electronics
This paper presents a novel compliant motion controller design for an omni-directional mobile robot. In this design, an external force observer is developed based on measuring motor current and speed without using an expensive force/torque sensor. The mobile robot has a handrail to assist the elderly to walk safely and stably. In this application, adaptive motion compliance is required in accordance with the applied force of the user. Practical experimental results show that the compliance of the walking helper can be adjusted by setting dynamic coefficients of the overall dynamical system. It was demonstrated that the external force observer successfully detected the pushing as well as pulling force of a user on the handrail. The velocity of the omni-directional walking helper is adjusted according to the inferred motion intent.
A New Design on Multi-Modal Robotic Focus Attention
(2008)
Lin, Chia-How
;
Yang, Chia-Hsing
;
Wang, Cheng-Kang
;
Song, Kai-Tai
;
Hu, Jwu-Sheng
Human detection and tracking is important for user-friendly human-robot interaction. The robot should be able to find the user autonomously and keep its attention to the user in a human-like manner. In this paper, a design and experimental study of robust human detection and tracking is presented through fusion several modalities of sensory information. The multi-modal interaction design utilizes a combination of visual, audio, and laser scanner data for reliable detection and tracking of an interested user. During tracking motion, obstacle avoidance behavior will be activated any time required to ensure safety. Furthermore, user can further assign the robot to interact with other user by speech command. Experimental results show that the robot can robustly tracks person under complex scenarios.
A new edge-adaptive demosaicing algorithm for color filter arrays
(2007-09-01)
Tsai, Chi-Yi
;
Song, Kai-Tai
;
電控工程研究所
;
Institute of Electrical and Control Engineering
A novel edge-adaptive demosaicing algorithm (EADA) is proposed in this paper to effectively reduce color artifacts in demosaiced images from a color filter array (CFA). The proposed algorithm aims to reduce the aliasing error in red and blue channels by exploiting high-frequency information of the green channel. To achieve this, color-difference based edge-adaptive filtering and post-processing schemes are designed to reproduce the color values by exploiting the green channel information. For green channel interpolation, any of the existing image interpolation methods can be used and combined with the proposed algorithm. Moreover, a new adaptive interpolation method is presented for reconstructing the green channel from CFA samples. We have compared this technique with two recently proposed demosaicing techniques: Gunturk's and Lu's methods. The experimental results show that EADA outperforms both of them in both PSNR values and CIELAB Delta E*(ab) measures. (C) 2007 Elsevier B.V. All rights reserved.
A new information fusion method for SVM-Based robotic audio-visual emotion recognition
(2007)
Han, Meng-Ju
;
Hsu, Jing-Huai
;
Song, Kai-Tai
;
Chang, Fuh-Yu
;
電控工程研究所
Emotion recognition has become an important research area for advanced human-robot interaction. Through recognizing facial expressions, a robot can interact with a person with a more friendly manner. In this paper, we proposed a bimodal emotion recognition system by combining image and speech information. A novel information fusion strategy is proposed to set proper weights to two feature modalities based on their recognition reliability. The fusion weights are determined by the distance between test data and the classification hyperplane and the standard deviation of training samples. After normalization using the mean distance between training samples and the hyperplane, the fusion weight is set to represent the classification reliability of individual modality. In the latter bimodal SVM classification, the recognition result with higher weight is selected. The complete procedure has been implemented in a DSP-based system to recognize five facial expressions on-line in real time. The experimental results show that a recognition rate of 86.9% is achieved, an improvement of 5% compared with using only image information.
A Novel Flag-Language Remote Control Design for a Laparoscopic Camera Holder Using Image Processing
(2015)
Zinchenko, Kateryna
;
Huang, Wayne Shin-Wei
;
Liu, Kai-Che
;
Song, Kai-Tai
;
電控工程研究所
Minimally invasive surgeries( MIS) possess obvious advantages for patients but require more specially trained personnel than common, open procedure. With modern advances in robotics, it becomes possible to ease these requirements by introducing robotized assistants into operation. However, for robot-assisted solo MIS, it is crucial to have appropriate methods for robot control. This work presents a design of flag language as a novel human-to-robot communication method based on image processing of the endoscope video during real-time intervention. The proposed system comprises autonomous positioning of endoscope holder combined with recognition of surgeon intention by analyzing surgical instrument postures. The proposed algorithm has been evaluated using experimental setup as well as video clips from laparoscopic operation. Experimental results show that the flag posture can be detected satisfactorily with speed of 12 frames per second, as well as system robustness under moderate lighting and noise conditions.
A Novel Sensory Mapping Design for Bipedal Walking on a Sloped Surface
(2012-10-12)
Wu, Chiao-Min
;
Huang, Chao-Ping
;
Hsieh, Chang-Hung
;
Song, Kai-Tai
;
電子工程學系及電子研究所
This paper presents an environment recognition method for bipedal robots using a time-delay neural network. For a robot to walk in a varying terrain, it is desirable that the robot can adapt to any environment encountered in real-time. This paper aims to develop a sensory mapping unit to recognize environment types from the input sensory data based on an artificial neural network approach. With the proposed sensory mapping design, a bipedal walking robot can obtain real-time environment information and select an appropriate walking pattern accordingly. Due to the time-dependent property of sensory data, the sensory mapping is realized by using a time-delay neural network. The sensory data of earlier time sequences combined with current sensory data are sent to the neural network. The proposed method has been implemented on the humanoid robot NAO for verification. Several interesting experiments were carried out to verify the effectiveness of the sensory mapping design. The mapping design is validated for the uphill, downhill and flat surface cases, where three types of environment can be recognized by the NAO robot online.
A Sensory Data Tracking Approach to Bipedal Gait Compensation Control on Slope Surfaces
(2013)
Wu, Chiao-Min
;
Song, Kai-Tai
;
電控工程研究所
;
Institute of Electrical and Control Engineering
The development of biped robots has gained much research attention in recent years due to their capacity and flexibility in providing assistance to human. In practical applications, these robots are expected to walk on various types of ground surfaces. Balance control of bipedal walking on uneven terrain is still a challenging problem. This paper proposes a novel gait compensating method using sensory data tracking such that a robot can keep balance and walk on slope surfaces. The real-time compensation control system works to adjust the gait and thus make online sensory data to track the stored target sensory data, which are obtained from the robot when it walks on a flat surface. While the robot adjusts its gait to make online sensory data similar to those on a flat surface, it will achieve a stable pose in walking. The proposed method has been tested for several slopes using the NAO robot. The robot can adjust its pose automatically on slope surface according to the proposed method. An experiment on transition slopes further validates that the method can be extended to more general terrain variations for biped walking.
A Study on Motion Control of a Robotic Endoscope Holder Using Speech Recognition
(2016)
Zinchenko, Kateryna
;
Wu, Chien-Yu
;
Song, Kai-Tai
;
資訊工程學系
;
電控工程研究所
This paper presents a design of speech recognition interface for a robotic endoscope holder for abdominal Minimally Invasive Surgery(MIS). A set of seven voice commands are used to control 3 degree of freedom(DOF) robotic arm with remote center of motion(RCM). Speech recognition algorithm was implemented on Ubuntu OS using Pocket Sphinx and achieved 90% success rate. Command signal was delivered to the robotic holder using serial port communication through RS232 cable. Experimental results show that the proposed control strategy has navigation precision up to 2.5mm or 7.7\'degrees in horizontal direction, 1.8mm or 3.150 degrees in vertical direction and 9 mm precision during insertion task with camera tip located at distance of 5 cm after RCM point. Average latency between two consecutive commands was 3.34s.
A Study on Speech Recognition Control for a Surgical Robot
(2017-04-01)
Zinchenko, Kateryna
;
Wu, Chien-Yu
;
Song, Kai-Tai
;
交大名義發表
;
電控工程研究所
Speech recognition is common in electronic appliances and personal services, but its use for industrial and medical purposes is rare because of the presence of motion ambiguity. For minimally invasive surgical robotic assistants, this ambiguity arises because the robotic motion is not calibrated to the camera images. This paper presents a design for a speech recognition interface for an HIWIN robotic endoscope holder. A new intentional speech control is proposed to control movement over long distances. To decrease ambiguity, a method is proposed for voice-to-motion calibration that compares the degree of change in the endoscope image for a voice command. A speech recognition algorithm is implemented on Ubuntu OS, using CMU Sphinx. The control signal is sent to the robot controller using serial-port communication through a RS232 cable. The experimental results show that the proposed intentional speech control strategy has a navigation precision of up to 3.1 degrees of angular displacement for the endoscope. The overall system processing time, including robotic motion, is 3.22 s for similar to 1.8-s speech duration. The reference image navigation range is from 2.5 mm for similar to 0.5- s speech duration up to 6 mm for similar to 1.8-s speech duration, using a setup with camera tip that is located at a distance of 5 cm from the remote center of motion point.
A visual circuit for lateral motion detection
(2007)
Song, Kai-Tai
;
Shen, Sung-Chih
;
電控工程研究所
;
Institute of Electrical and Control Engineering
This paper presents a design and implementation of an intelligent vision chip for detecting lateral object motion. We propose a correlation interconnection scheme for the outputs of a photocircuit array to infer motion direction of nearby objects as well as the tendency of approaching of a moving object. This design facilitates lateral obstacle avoidance of an intelligent vehicle. The circuit has been successfully simulated by Hspice and integrated into a sensor chip of 20x20 pixel array. The vision chip has been fabricated with TSMC 0.35 mu m 2P4M CMOS process. The chip area is 2.48 mm x2.42 mm. The function of photocells, registers and correlation in extreme cases are verified.
An Adaptive Routing Protocol for Health Monitoring with a Sensor Network and Mobile Robot
(2010)
Yang, Su-Hen
;
Song, Kai-Tai
;
電機工程學系
;
Department of Electrical and Computer Engineering
In this paper, we developed an adaptive route selection algorithm for a Mobile Ad-hoc Network (MANET). This algorithm has been realized and implemented for health monitoring applications with sensor nodes and a mobile robot. The health care devices include a body-pose estimation module, and a SpO(2) and ECG sensor module. These sensor modules wore on individuals and the mobile robot itself form a mobile sensor network. Conventional Ad-hoc On-demand Distance Vector (AODV) looks for a route which has the smallest hop count, but the route with the smallest hop count might not be the best route. In this paper we propose a novel routing protocol that features to combine signal strength and battery level into account, and uses fuzzy inference to search for a route with stable RF signals. Simulation results show that the packet delivery ratio increases from 81% to 90%. In practical experiment on a Zigbee MANET, the packet delivery ratio improves from 66% to 94%. The proposed MANET has also been experimented in practical health care application by integrating a pose estimation module, a SpO(2) sensor and a mobile robot.
AN INTERACTIVE CONTROL ARCHITECTURE FOR MOBILE ROBOTS
(2013)
Lin, Chia-How
;
Song, Kai-Tai
;
電子工程學系及電子研究所
;
電控工程研究所
;
交大研究總中心
Programming a robot to handle daily-life tasks in a dynamic changing environment with versatile and responsive behaviours is challenging. An intelligent robot needs to integrate many heterogeneous modules of various features using a powerful control architecture. This study aims to develop a methodology for designing a flexible real-time robot control system using an agent-based approach. The proposed control architecture decomposes a complex robotic control system into agents and allows heterogeneous modules to he integrated more efficiently. The software framework solves resource sharing, message flow and coding problems, and helps to integrate separate processes into a whole system with less effort. The proposed core components guarantee real-time response and high performance of the robotic system. This approach is also suitable for system integration of a research team, where the software modules from different, developers can be integrated on the robot and teamwork can he realized much easier. Interesting experiments on robotic focus attention and human robot interaction using two laboratory-developed robots demonstrate the effectiveness of the proposed control architecture.
Autonomous and Stable Tracking of Endoscope Instrument Tools with Monocular Camera
(2012)
Song, Kai-Tai
;
Chen, Chun-Ju
;
電機工程學系
;
Department of Electrical and Computer Engineering
With an autonomous tool tracking system, surgeons do not need to handle the endoscope during surgery. This paper aims to develop an image tracking system with a stable view for minimally invasive surgery (MIS) by using monocular endoscope. In order to provide a stable view for the surgery, we propose to set a buffer zone in the center of image frame for the robotic camera holder to track the endoscope instruments. If endoscope instruments are inside the buffer zone, the endoscope robot will keep still. However, if one or both of them move away from the buffer zone, the robot will start to track the instrument. Furthermore, by calculating the distance between two instruments in the image plane, the robotic camera holder can track the depth to the instruments. The robot will therefore zoom in if the distance is too short. A safe distance from the camera lens to instrument tools can therefore be maintained during the image tracking. Preliminary experimental results show that the proposed design achieves satisfactory tracking accuracy in real time and provides a stable view simultaneously.