Handling Complex Smart Walker Interaction Strategies with Behavior Trees
Nombre: PAULA ALCANTARA CARDOSO
Fecha de publicación: 24/10/2022
Supervisor:
Nombre |
Papel |
|---|---|
| ANSELMO FRIZERA NETO | Advisor * |
| RICARDO CARMINATI DE MELLO | Co-advisor * |
Junta de examinadores:
| Nombre |
Papel |
|---|---|
| ANSELMO FRIZERA NETO | Advisor * |
| EDUARDO ROCON DE LIMA | External Examiner * |
| PABLO JAVIER ALSINA | External Examiner * |
| RICARDO CARMINATI DE MELLO | Co advisor * |
Sumario: Walking is an activity that requires mastering stability and precision in order to be learned. Between the ages of 8 and 10, gait becomes an automatic activity, which once mastered does not depend on the attention of individuals. Mechanical factors and diseases, such as Parkinsons and spinal cord injuries, as well as physical and cognitive conditions, can contribute to the quality of gait, which starts to present different types of disorders. With the increase of life expectancy and the share of the population over 65 years old, there is a concern regarding the demands for accessibility, rehabilitation and assistance, since they suffer from the reduction of their balance and mobility capacities. In this context, mobility assistance devices are valuable options to meet the needs of their users. Walkers, in particular, are alternatives for those who have residual motor skills. Its adoption postpones the use of wheelchairs and encourages the independence of its users. The inclusion of technologies, such as sensors and actuators, in the structure of walkers makes them intelligent, as they allow the incorporation of new functionalities that provide greater
and better assistance for locomotion. The devices called robotic or smart walkers have several modules and control strategies that make their operation - and understanding - complex. This Masters Dissertation proposes the inclusion of an artificial intelligence algorithm, based on a hierarchical architecture, for decision making that is capable of integrating several control strategies human-robot-environment interactions in the UFES CloudWalker. The algorithm implemented was the Behavior Tree, a structure that allows switching between controllers in a modular and reactive way. The system was validated by volunteers who performed a series of tasks aimed at evaluating the global performance of the smart walker. As a result, the system proved to be able to handle complex interactions between user, walker and the environment during navigation.
Keywords: Behavior Trees, Human-robot-environment Interaction, Smart Walker, Artificial Intelligence.
