Publicação
Development of an intelligent robotic system for rehabilitation of upper limbs using a collaborative robot
| Resumo: | Rehabilitation is a relevant process for the recovery from dysfunctions and improves the realisation of patient’s Activities ofDaily Living (ADLs). Therefore, the development of technologies for this field has significant importance because the improvement of the rehabilitation can affect many people. This work proposes a robotic system for the rehabilitation of the upper limbs using a collaborative robot and an intelligent control algorithm that makes the solution robust and adaptable to each patient. The UR3 from Universal Robots© was used to implement two Reinforcement Learning algorithms, the SARSA and Q-learning, applied to this rehabilitation problem. The goal of this system provides a common training force applying resistance on the movement performed by the patient. This thesis is divided into twomain parts. The first one was the development of a simulation composed by the UR3 and a human model in V-REP platformthat could be controlled through a dedicated interface or externally through the MATLAB using the self-control algorithms. This simulation was created with a graphical interface for visualisation, and a human-machine interface, to control the robotic system with RL algorithm, built onMATLAB. The results obtained with the simulation presented the expected system behaviour. The second part was the experiment of the real system with a healthy subject. The experiment was divided in two phases the first considering the training only in one axis and second in the three Cartesian axes. The used algorithms were the same as the simulation, but in this case, they were implemented in Python language. The experiment considering one axis presents satisfactory results, while for the three axes the results were not so good. The obtained results with the real system experiment for one and three axis were compared with the human armmodel proposed in other studies to validate the applied methodology. This work represents an important contribution for the field because presents a new feature to help therapists and patients to get better results in the rehabilitation process. |
|---|---|
| Autores principais: | Fernandes, Lucas de Azevedo |
| Assunto: | Reinforcement learning Robotic rehabilitation Human arm model Collaborative robots Upper limbs rehabilitation |
| Ano: | 2021 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Instituto Politécnico de Bragança |
| Idioma: | inglês |
| Origem: | Biblioteca Digital do IPB |
| Resumo: | Rehabilitation is a relevant process for the recovery from dysfunctions and improves the realisation of patient’s Activities ofDaily Living (ADLs). Therefore, the development of technologies for this field has significant importance because the improvement of the rehabilitation can affect many people. This work proposes a robotic system for the rehabilitation of the upper limbs using a collaborative robot and an intelligent control algorithm that makes the solution robust and adaptable to each patient. The UR3 from Universal Robots© was used to implement two Reinforcement Learning algorithms, the SARSA and Q-learning, applied to this rehabilitation problem. The goal of this system provides a common training force applying resistance on the movement performed by the patient. This thesis is divided into twomain parts. The first one was the development of a simulation composed by the UR3 and a human model in V-REP platformthat could be controlled through a dedicated interface or externally through the MATLAB using the self-control algorithms. This simulation was created with a graphical interface for visualisation, and a human-machine interface, to control the robotic system with RL algorithm, built onMATLAB. The results obtained with the simulation presented the expected system behaviour. The second part was the experiment of the real system with a healthy subject. The experiment was divided in two phases the first considering the training only in one axis and second in the three Cartesian axes. The used algorithms were the same as the simulation, but in this case, they were implemented in Python language. The experiment considering one axis presents satisfactory results, while for the three axes the results were not so good. The obtained results with the real system experiment for one and three axis were compared with the human armmodel proposed in other studies to validate the applied methodology. This work represents an important contribution for the field because presents a new feature to help therapists and patients to get better results in the rehabilitation process. |
|---|