Abstract
So far, many different robots have been built and developed to help people, improve the quality of life or make work easier. Nowadays we live in a society where pressure arises due to work and other factors, causing people to develop stress in their lives. A massage can have a relaxing effect and help alleviate stress, such as a foot massage.This thesis presents a prototype of a vision-based foot massage robot. For this purpose,a hardware and a software setup was created. One part of the hardware setup was themodification of the robotic arm. This involved replacing the claw, which originally served as an end-effector, with a deodorant roller. This deodorant roller is additionally equipped with five buttons, which are supposed to detect the physical contact between the robot and the sole. Furthermore, a platform and a footrest was built.One of the main goals of the software setup is to identify and recognize the sole of thefoot as well as the robot. To achieve this goal, a Mask Region-Based Convolutional Neural Network was trained. This network marks the pixels from the robot and the sole of the foot on the input image. The massage points are manually generated once on a sole mask template and projected on to the mask of the foot during the massage. In addition, these points are loaded into a simulation, where the robotic arm is also mapped into. The rotation angles of each servomotor are calculated in the simulation with inversekinematics, so that the end-effector reaches the massage point. Then these angles aresent to the robot controller, in this case an Arduino, to move the robot. If physical contact is detected, the massage technique frictioning will be performed. Furthermore,some settings of the massage movements, such as the speed, can be changed. For this purpose, three Convolutional Neural Networks were trained, which are used for audio command recognition.Finally, the prototype was tested in a user study with a group of 13 participants. Through the evaluation it was discovered that a robotic arm can perform a state-of-the-art footmassage. The stress level of the participants is lower than before the massage and they reported to be more relaxed. However the robot massages the right foot more accurate than the left one and the robot does not always respond well to the voice commands.
Reference
Ammer, D. (2023). Development of a vision-based foot massage robot [Diploma Thesis, Technische Universität Wien]. reposiTUm. https://doi.org/10.34726/hss.2023.97160