Li, Qinbo (2016-08). An Intelligent Tutoring System for Computer Numerical Control. Master's Thesis. Thesis uri icon

abstract

  • In recent years, the use of Intelligent Tutoring Systems (ITS) in classrooms and communities has increased and they proved to be very effective. For the domain of Computer Numerical Control (CNC), however, existing approaches in ITS are not applicable or will not work well. CNC programming is different from computer programming languages, and students fail to solve CNC programming problems mainly due to two reasons: (1) lack of problem solving skills and (2) misconceptions or missing facts. CNC programming requires that students master a lot of facts and concepts before they try to write a program. We built an ITS for CNC called the "CNC-Tutor" and proposed a data-driven approach that can generate proper hints and feedback during the students' problem solving process. This approach is based on finding the most similar past submissions with the current student's solution. The similarity is measured by the proposed "Behavior & Machine state distance" metric. Experiments show that the generated hints can help the students solve the CNC programming problem and the generated feedback can help the students to find their misconceptions. A survey on the effectiveness of our CNC-Tutor shows a positive impact on the students.
  • In recent years, the use of Intelligent Tutoring Systems (ITS) in classrooms and communities has increased and they proved to be very effective. For the domain of Computer Numerical Control (CNC), however, existing approaches in ITS are not applicable or will not work well. CNC programming is different from computer programming languages, and students fail to solve CNC programming problems mainly due to two reasons: (1) lack of problem solving skills and (2) misconceptions or missing facts. CNC programming requires that students master a lot of facts and concepts before they try to write a program.

    We built an ITS for CNC called the "CNC-Tutor" and proposed a data-driven approach that can generate proper hints and feedback during the students' problem solving process. This approach is based on finding the most similar past submissions with the current student's solution. The similarity is measured by the proposed "Behavior & Machine state distance" metric. Experiments show that the generated hints can help the students solve the CNC programming problem and the generated feedback can help the students to find their misconceptions. A survey on the effectiveness of our CNC-Tutor shows a positive impact on the students.

publication date

  • August 2016