Abstract
One-on-one student-instructor communication is essential in many aspects for mechanical engineering education, like personal instruction on computer programming, and hands-on skills training in robotic assembly. Especially under the effect of COVID-19, individual student-instructor interactivity becomes increasingly precious yet growingly difficult. To address this issue, we aim to build a so-called “cloud teacher”, which is a textual-based conversational agent for answering the questions from students in Mechanical Engineering via machine learning technologies. The basic underlying idea is to train the agent by utilizing open-source artificial intelligence tools from Google’s DeepMind, such that the agent can understand and answer the questions raised by students with an acceptable confidence value. Specifically, it is expected to customize a training database combining (1) the U.S. Mechanical Engineering Syllabus for Engineering Graduates; (2) Functions, operators, and logics in Arduino and SolidWorks programming, two most common software in robot control and 3D printing; and (3) VEX hardware assembling information. In addition, the answers with associated confidence value will be reviewed by a panel of teachers in Mechanical Engineering while further training on the agent could be launched based on review results. Moreover, Big Data Analytics can be conducted accordingly based on students’ historical questions, and prescribed teaching can be provided by the teacher for individualized instruction on each student or focusing instruction for some common questions raised by students. An Android-based mobile application is expected to be produced. As a result, students can raise their questions conveniently and get instantaneous feedback with a 24-7 service. Meanwhile, individualized instruction from course teachers could be provided for each student regarding the result of Big Data Analytics on students’ historical questions. The method will demonstrate its effectiveness in 4 existing courses (MAEG1010 Introduction to Robot Design, MAEG2601 Technology, Society and Engineering Practice, MAEG3920 Engineering Design and Applications, and UGEB2303 Robots in Action).
Brief write-up
Project objectives
1. Build a “cloud teacher”, which is a textual-based conversational agent for answering the questions from students in Mechanical and Automation Engineering via machine learning technologies.
2. Develop a set of Python programming codes for the “cloud teacher” autonomously learning in the area of VEX hardware assembling, Arduino and Solidworks programming.
3. Conduct Big Data Analytics based on students’ historical questions, and provide prescribed teaching with individualized instruction on each student or provide focusing instruction for some common questions raised by students.
4. Develop a mobile APP with the intelligent conversational agent and implement it in 4 courses, i.e., MAEG1010 “Introduction to Robot Design”, MAEG2601 “Technology, Society and Engineering Practice”, MAEG3920 “Engineering Design and Applications”, and UGEB2303 “Robots in Action”.
5. Make 10 sets of remote controllable robotic arms by students with the help of proposed APP.
Activities, process and outcomes
The whole process of the project development can be generally divided into four parts: The first part focuses on training the intelligent agent with a database of fundamental knowledge of Mechanical Engineering related to the U.S. Mechanical Engineering Syllabus for Mechanical Engineering graduates. The second part is concerned with the training of the intelligent agent regards to a database of Arduino programming. The third part aims to train the intelligent agent with a database of VEX hardware assembling. In the last part, a mobile APP will be developed and implemented in an online robotic laboratory.
Deliverables and evaluation
Summary of project deliverables: (a) 1 set of Python programming codes developed for training the smart agent; (b) 1 mobile APP developed and implementation on 1 online robotic laboratory; (c) 2 presentations at local conference; (d) 2 local seminars; (e) 1 leaflet; and (f) 10 sets of remote controllable robotic arms.
Summary of evaluation methods for this project: (a) survey on the online laboratory learning experience towards the end of 4 courses; (b) survey on the user-experience on the conversational intelligent agent towards the end of 4 courses; (c) focus group interview with a small group of volunteer students of 4 courses; and (d) feedbacks at seminars and conferences.
Dissemination, diffusion, impact and sharing of good practices
Summary of dissemination: 2 presentations in 2 seminars, 2 presentations in 2 local conferences.
Impact: Collaboration with an oversea University Cardiff Metropolitan University and got one external grant. Promoted by the Faculty of Engineering and got one internal extended grant. Replication in Faculty foundation courses and department elective courses.
Impact on teaching and learning
The rate of positive feedbacks of targeted courses are all above 70% in both the survey on the online laboratory learning experience and also the survey on the developed cloud teacher. The performance of students in some concerned assessments have been remarkably elevated in some targeted courses.
