Syntéza plánů pro vzdělávání osob s poruchou autistického spektra s pomocí robotů
Název práce v češtině: | Syntéza plánů pro vzdělávání osob s poruchou autistického spektra s pomocí robotů |
---|---|
Název v anglickém jazyce: | Plan synthesis for robot-assisted education of people with autism spectrum disorder |
Klíčová slova: | plánování|strojové učení|robotika|výuka |
Klíčová slova anglicky: | planning|machine learning|robotics|teaching |
Akademický rok vypsání: | 2024/2025 |
Typ práce: | disertační práce |
Jazyk práce: | |
Ústav: | Katedra teoretické informatiky a matematické logiky (32-KTIML) |
Vedoucí / školitel: | Stefan Edelkamp, Dr. rer. nat. |
Řešitel: | skrytý![]() |
Datum přihlášení: | 08.10.2024 |
Datum zadání: | 08.10.2024 |
Datum potvrzení stud. oddělením: | 08.10.2024 |
Konzultanti: | RNDr. David Obdržálek, Ph.D. |
Zásady pro vypracování |
The student should create a computerized system for the automatic assessment and planning of structured tasks in education.
This system can be used in the education of an autistic person both for preparing the educational task and then using robot assistence as a teacher. Robot as teacher: The robot assists the person in completing a simple task (by manually moving elements, pointing, verbal cues, ...). The person can ask the robot for help (e.g. by using a button, gesture, voice command, ...). There are many interaction strategies (what the robot should do and when) - to avoid the psychological aspect, multiple strategies can be implemented, as options for the user. Planned actions are translated into abstract robot actions using the selected strategy; abstract robot actions are translated into commands for the real robot (Nao, UR, ...). |
Seznam odborné literatury |
Pérez-Vázquez, E., Lorenzo, G., Lorenzo-Lledó, A. et al. Analysis of the Application of the Bee-Bot Robot for the Development of Social Reciprocity
Skills in Students with Autism Spectrum Disorder. Int J of Soc Robotics (2024). https://doi.org/10.1007/ Katsanis, I.A.; Moulianitis, V.C.; Panagiotarakos, D.T. Design, Development, and a Pilot Study of a Low-Cost Robot for Child–Robot Interaction in Autism Interventions. Multimodal Technol. Interact. 2022, 6, 43. https://doi.org/10.3390/mti6060043 A. Z. Amat et al., "Design of a Desktop Virtual Reality-Based Collaborative Activities Simulator (ViRCAS) to Support Teamwork in Workplace Settings for Autistic Adults," in IEEE Transactions on Neural Systems and Rehabilitation Engineering, vol. 31, pp. 2184-2194, 2023, doi: 10.1109/TNSRE.2023.3271139. S. G. T. Haokip, G. Shah and U. Lahiri, "Psycho-physiological implications of computer based social and non-social interactive tasks for children with autism," 2017 8th International Conference on Computing, Communication and Networking Technologies (ICCCNT), Delhi, India, 2017, pp. 1- 7, doi: 10.1109/ICCCNT.2017.8204153. Mechling, L. C., & Swindle, C. O. (2013). Fine and Gross Motor Task Performance When Using Computer-Based Video Models by Students With Autism and Moderate Intellectual Disability. The Journal of Special Education, 47(3), 135-147. https://doi.org/10.1177/0022466911433859 Baraka, K., Melo, F.S., Couto, M. et al. Optimal action sequence generation for assistive agents in fixed horizon tasks. Auton Agent Multi-Agent Syst 34, 33 (2020). https://doi.org/10.1007/s10458-020-09458-7 |
Předběžná náplň práce |
Studies have shown that children with autism often prefer a robot as a teacher and that robot assistance leads to significant improvements in their skills.
The shoebox tasks are part of a teaching approach developed by TEACCH® Autism Program (University of North Carolina). TEACCH is the recommended methodology in the Czech Republic and is used worldwide. The TEACCH Structured Work Session consists of "structured tasks". One shoebox task is a structured task with a box that serves as a workstation and contains all the materials for the task (puzzles, cards, blocks, ...). There is a visual structure that allows to recognize how the task should be completed. In general, the task can be described and solved using logic programming. Then, for a given task instance, a vision-based method can be used to decide whether the task has been completed or what state it is in. Another approach might be to try to infer the goal using machine learning methods e.g., using deep learning from annotated samples of completed and incomplete instances (images or simulation snipplets) and to create a logic model from the visual representation. Planning and plan synthesis would use some predefined actions to manipulate objects. Possible extensions: Automatically generate 3D models of shoebox tasks in Gazebo/Unity/MuJoCo/Issak (to provide the visual feedback or to practice the task in a simulation). Use vision language model to generate verbal instructions. |