• Chih-Hung Gilbert Li



    National Taipei University of Technology (Taipei Tech)


    Graduate Institute of Manufacturing Technology


    Industry 4.0 Laboratory


  • Personal Information


    Education 學歷

    Ph.D. Carnegie Mellon University / Mechanical Engineering


    M.S. Carnegie Mellon University / Mechanical Engineering


    B.S. National Tsing Hua University / Power Mechanical Engineering


    Taipei Municipal Jianguo High School


    Experience 經歷

    Associate Dean / College of Mechanical and Electrical Engineering / National Taipei University of Technology


    Associate Professor / National Taipei University of Technology


    Director / Office of International Affairs / Minghsin University of Science and Technology


    Director / Automated Vehicles and Equipment Development Center


    Engineering Specialist / Lord Corporation (USA)


  • Fields 領域

    務實 創新 合作 堅持

    Research of Industry 4.0


    Artificial Intelligence, Cyber Physical System, Internet of Things, intelligent robots and vehicles


    Industry 4.0 is a collective noun. Its technologies such as Internet of Things, big data, cloud computing, artificial intelligence, automation, etc. are revolutionizing many industries including manufacturing. It is expected that not only will much of the production and management efficiency and flexibility be significantly increased, but Industry 4.0 is also more likely to change many existing commercial and industrial operating models. Through systematic research and testing, we are committed to proposing forward-looking and innovative operating models or technologies, such as the development of intelligent service robots, related topics of human-machine collaboration, the automated personal rapid transit system, etc. to promote the advancement of technology for human well-being.


    Structural Stress Analysis (Finite Element Analysis)


    Structural topology optimization, nonlinear stress and strain analysis, fatigue and fracture analysis


    We have accumulated more than 20 years of experience in the finite element analysis. Varieties of linear or nonlinear structural stress problems were solved using the finite element software such as ANSYS. Projects include simple models such as trusses or elastic structures and more complex ones such as huge composite structures, large deformation or high strain analysis, contact and friction analysis, plastic deformation analysis, fatigue and fracture analysis, and dynamic collision analysis. In addition, by using the ANSYS APDL, projects that require large amounts of finite element analyses can be efficiently processed and completed. In the advanced design, the topology optimization design of the structure is obtained by using the artificial intelligence algorithm or the Evolutionary Structural Optimization method.

    累積個人與團隊超過20年的有限元素分析技術。舉凡各種線性及非線性之結構應力分析,透過有限元素分析軟體(如ANSYS)的運用,都可以迎刃而解。較簡單的如衍架分析與彈性結構分析等。較複雜的有超大型複合結構分析、彈性體大變形或大應變分析、接觸與摩擦分析、塑性變形分析、疲勞與破裂分析、及動態碰撞分析等。此外,透過程式自動化規劃(APDL),可以高效率處理需要大量有限元素分析的專案。在進階設計方面,運用人工智慧演算法或進化式結構拓樸最佳化法 (Evolutionary Structural Optimization)獲得結構之拓樸最佳化設計。

    Development of Innovative Mechanisms


    Guitar robot, monorail system, mechanical damper, electromagnetic actuator, bus sliding door, vehicle suspension, retractable carriage, integrated music sounding teaching device, etc.


    We are committed to invention and design of patent-protected mechanisms. Previous industry-university cooperation research projects include, but are not limited to, artificial intelligence applications, smart robots, intelligent automated transport systems, novel actuators, robot mechanisms, innovative shock absorbers, various mechanical structures, and equipment with special functions for vehicles. More than 20 domestic or foreign patents have been obtained, and many have been authorized to the industry.


  • Projects 實績


    Holistic-view Deep Learning for Automatic Windshield Wiper Activation


    Deep Learning Application


    A windshield rain detection system using holistic-view deep learning is constructed in this project. A wiper control algorithm based on a time-series treatment is also presented. The video images of ordinary driving recorders were used to train a deep convolutional neural network for wiper activation classification. Overall, we achieved an average precision rate of 0.88 in our video-based rain detection experiments; our recall rate of 0.87 is significantly higher than the state-of-the-arts that averaged around 0.6. It is also proved that the proposed system is practical for real-time vehicle windshield rain detection and wiper control. In this film, a blue square indicates that our detection system recommends that the wiper should activate, and a yellow circle indicates it should not.

    在本專案中我們構建了採用整體視覺深度學習的擋風玻璃雨水探測系統,還提出了一種基於時間序列處理的雨刷控制算法。我們使用普通行車記錄儀的視頻圖像以訓練深度卷積神經網絡,以進行雨刷啟動分類。總體而言,我們在基於視頻的雨水檢測實驗中實現了0.88的平均精確率; 我們的召回率達到0.87,亦明顯高於平均約為0.6的現有一般水平。我們所提出的系統亦證明了實時車輛擋風玻璃雨水檢測和雨刷控制是可行的。在此影片中,藍色方塊表示我們的偵測系統建議雨刷應作動,而黃色圓圈表示雨刷不應作動。

    Chi-Cheng Lai, Chih-Hung G. Li*, "Video-Based Windshield Rain Detection and Wiper Control Using Holistic-View Deep Learning," in Proc. 2019 IEEE 15th International Conference on Automation Science and Engineering (CASE 2019) Vancouver, BC, Canada.

    Indoor Self Driving of Balancing Robot


    Deep Learning Application


    In this project, the team uses a depth camera to capture continuous images of the front scene while the balance car is moving fast indoors. Learning through the deep convolutional neural networks, robots are taught to react naturally like humans while encountering various conditions in the indoor environment, such as obstacles, left and right walls, moving objects, and so on. The red indicator in the video represents the direction response of the deep neural network by instant image recognition. The reaction can be further used to control the electromechanical system of the robot to controll the direction of movement of the robot.



    Indoor Place Recognition of Mobile Robot


    Deep Learning Application


    We built the indoor localization capability of mobile robots with a purely visual architecture. By using a plurality of cameras mounted on the robot and capturing images at multiple predetermined positions along the path, visual feature training sets were established and used to train a location classifier. Using deep learning architecture, we train the robot to recognize the global features of each position. In this test film, one can see that the robot recognizes each location when it navigates along the corridor. The code displayed at the upper left corner of the video changes from 0 to 20 in order. The overall precision rate is 92%; the recall rate is 87%.


    Yi-Feng Hong, Yu-Ming Chang, Chih-Hung G. Li*, "Real-time Visual-Based Localization for Mobile Robot Using Structured-View Deep Learning," in Proc. 2019 IEEE 15th International Conference on Automation Science and Engineering (CASE 2019) Vancouver, BC, Canada.

    Intelligent Robot Motion Control


    Deep Learning Application


    Mobile service robots can provide many functional services such as personal guidance, delivery, cleaning, transportation, equipment operation, assembly line assistance, security, medical assistance, etc. To be able to perform the above services, the robots' ability to detect, maneuver, and work in the environment is crucial. Our laboratory uses the latest deep learning ConvNet image recognition technology to enhance the robot's working ability. Through the vision-position direct control method developed by the team, the robot directly obtains coordinate information from the observed images for motion control. It can significantly shorten the robot's training time in the workplace. Convolutional neural networks have a high degree of abstraction and classification capability, and can deal with many environmental variations or signal noises. They exhibit stronger and more intelligent identification capabilities than the traditional image processing methods. They are very suitable for applications in the fields of manufacturing, medical care, service, transportation, etc.


    Yu-Ming Chang, Chih-Hung G. Li*, Yi-Feng Hong, "Real-Time Object Coordinate Detection and Manipulator Control Using Rigidly Trained Convolutional Neural Networks," in Proc. 2019 IEEE 15th International Conference on Automation Science and Engineering (CASE 2019) Vancouver, BC, Canada.

    Real-time Object Coordinate Detection Using Deep Learning


    Deep Learning Application


    We use a deep learning framework for training object coordinate detection based on a single basis photo. As shown in the video, to train a coordinate detection scheme of a specific object, it only takes about 4 minutes on an ordinary notebook PC from taking the basis photo to completing the deep learning process.


    Workpiece Visual Placement Using Deep Learning


    Deep Learning Application


    The deep learning framework for object coordinate detection can be applied to the precision visual placement of a workpiece. In the video, the robot arm is installed on a mobile platform. The deep learning visual detection system endows the manipulator a placement precision of less than 1 mm.


    Industry 4.0 introduction method for small and medium enterprises


    Can SMEs introduce the technologies of Industry 4.0, implement production tracking, flow-control on raw materials and finished products, and collect big data to optimize production processes or the product itself? The answer is - YES! By slightly expanding your basic ERP/MES and adopting a low-cost QR Code system, we can easily implement an automatic tracking system on the flow of raw material, storage, work order, scheduling, and shipment. The important parameters in the manufacturing processes can also be captured with sensors for real-time monitoring. The accumulated data can be stored to establish a production history database. The long-term accumulated data can be further analyzed using the Artificial Intelligence to optimize the production process or the products' performance and quality.


    答案是-YES! 透過稍微擴充基礎的ERP/MES系統,就可以低成本之二維碼掃描系統將原料進貨、倉儲、工單領料、排程、出貨等作業制度化。製程中之重要參數則可加裝感測器進行實時監控,並將數據上傳與儲存,建立生產履歷資料庫。長期累積的數據則可更進一步以各種人工智慧的方法進行分析,以優化製程或產品性能與品質。

    Development of educational materials

    for Industry 4.0 and Artificial Intelligence




    Development of intelligent and automated personal rapid transit (PRT)






    Chih-Hung Li*, Zong Jun Lu, "An Innovative Straddle Monorail Track Switch Design for the Personal Rapid Transit," International Journal of Heavy Vehicle Systems (SCI), 2019.

    Innovative Soft Actuator



    In this research project, we have successfully developed an innovative electromagnetic actuator, which not only has the characteristics of quietness and softness, but also has the features of simple structure, moderate power, easy control, and low cost, ideal for applications such as the robots and automated machines that require superior quietness or human-machine collaboration. The specially designed tapered elastomer provides a highly nonlinear elastic response that achieves force equilibrium with the highly nonlinear electromagnetic force at various displacements and voltages. This actuator has the special characteristics that the input voltage is linearly proportional to the output displacement, and thus the motion control can be easily performed using a simple open circuit.


    C. G. Li and H. P. Nguyen, “Development of a linearly responsive electromagnetic actuator,” presented at Int. Conf. Computer Science, Data Mining & Mechanical Eng., Bangkok, Thailand, Apr. 20–21, 2015.

    Quiet guitar robot



    In order to solve the noise problems often associated with robots or automation equipment, we developed an innovative silent electromagnetic actuator. In addition to providing silent linear actuation, simple control of linear voltage response was also achieved. In this project, a guitar robot was created; the experimental evidence has shown that the mechanical noise of the guitar robot is much lower than that of conventional actuators such as pneumatic cylinders, servo motors, stepping motors, solenoids, et al., and is much lower than the guitar sound itself.


    Chih-Hung G. Li*, Ming-Chang Lin, Basil A. Bautista, and Bettina E. To, "A Low-Noise Guitar Robot Featuring a New Class of Silent Actuators," IEEE ASME Transactions on Mechatronics (SCI), 2019.

    C. G. Li and B. P. Bautista, “On the compression of a stack of truncated elastomeric cones as a nonlinearly responsive spring,” Mech. Res. Commun, vol. 69, pp. 146–149 (SCI), 2015.

    中華民國發明專利 / 可撥弦之機械手指裝置 / 發明人李志鴻、包提達巴希爾 / 2017 / I582752

    Design optimization for monorail chassis structure




    Li CG. Design of the lower chassis of a monorail personal rapid transit (MPRT) car using the evolutionary structural optimization (ESO) method. Structural and Multidisciplinary Optimization, 54 (1): 165-175 (SCI), 2016.

    Analysis and testing of huge equipment in amusement parks



    • 代表作品1:本案對摩天輪模型進行高速風洞測試,量測模型在各級風速下之受力,以預測真實的120米摩天輪是否能承受預定之風力。本實驗室製作一座二百分之一的摩天輪模型置於風洞中接受試驗,並透過相似性理論推測出實體摩天輪所可能承受之力量。最後用有限元素分析估算在各級風力下摩天輪主結構所承受之應力程度,並確認摩天輪主結構之強度。
    • 代表作品2:超大型遊戲設備飛行平台具有六軸自由度,且極限操作項目達三百餘種,再加上其架構非常複雜,因此各部位結構之應力與疲勞分析難度相當高。本專案以有限元素分析軟體建構擁有一千七百萬元素之模型,並透過自行撰寫之分析自動化程式,使得本分析得以在短時間內完成。

    Patented automobile suspension strut featuring constant frequency




    Li CG. A novel suspension strut featuring constant resonance frequency. International Journal of Heavy Vehicle Systems, 22 (4): 293-310 (SCI), 2015.

    Fatigue analysis of high speed pump shaft高速幫浦轉軸疲勞分析



    Patented retractable carriage design




    中華民國新型專利 / 可伸縮之車廂機構 / 發明人李志鴻 / 2015 / M496590

    Improvement on the brim profile of thick spin-coating layers




    Patented bus slide door




    美國發明專利 / Longitudinal-Slide Door Controlling Mechanism /發明人Chih-Hung Li / 2012 / US 8292349

    中華民國新型專利 / 巴士之橫移式門體連動機構 / 發明人李志鴻 / 2011 / M418828

    Patented lub-rubber dampers




    美國發明專利 / Cabinet Door Buffer Bar / 發明人Chih-Hung Li / US 7076834

    中華民國發明專利 / 緩衝棒 / 發明人李志鴻 / 2003 / 538202

    中華民國發明專利 / 櫃門緩衝棒 / 發明人李志鴻 / 2004 / I225533

    Integrated music sounding teaching device




    中華民國新型專利 / 整合式音樂發聲教學裝置 / 發明人廖美瑩、林鈺姍、李志鴻 / 2015 / M515186


    Dynamically Balanced Robot with a Manipulator


  • Contact


    Welcome to contact us regarding your need for collaborative R&D or engineering service.



    台北市忠孝東路三段一號 綜科館714-2室
    +886-2-2771-2171 ext. 2092