Desarrollo de un controlador de posición avanzado para endoscopio blando en cirugía laparoscópica
No hay miniatura disponible
Fecha
2023-11-17
Autores
Título de la revista
ISSN de la revista
Título del volumen
Editor
Pontificia Universidad Católica del Perú
DOI
Resumen
El presente estudio se desarrolla en el marco de brindar asistencia al cirujano en la
laparoscopia, la cual es una cirugía utilizada para tratar problemas de salud en la zona
abdominal. El procedimiento utiliza una cámara conectada a un tubo delgado flexible llamado
endoscopio, el cual permite observar al interior de la zona abdominal del paciente; las
imágenes obtenidas por el instrumento son utilizadas por el cirujano durante el tratamiento
del paciente. Para garantizar un procedimiento correcto, se debe mover correctamente el
endoscopio en el interior del abdomen, siendo esta tarea específicamente la que se busca
facilitar su control y con ello dar apertura a una serie de posibilidades como el movimiento
asistido, la operación remota y la automatización completa de la tarea.
Con el fin de proponer una solución, actualmente, con el avance en el campo de la robótica
blanda se han diseñado y fabricados manipuladores o actuadores blandos que puedan ser
usados como endoscopios, los cuales tienen la capacidad de deformarse y ser forzados a
moverse para alcanzar diferentes posiciones deseadas dentro de sus límites de operación.
El cuerpo del manipulador o actuador blando en estudio presenta cuatro cámaras internas,
las cuales pueden ser deformadas regulando la cantidad de presión de aire al interior de cada
cámara. Para controlar y alcanzar la posición deseada del efector final del endoscopio, donde
una cámara será conectada, en el presente trabajo se realiza el modelamiento de la dinámica
del cuerpo del endoscopio y el diseño del controlador de posición.
La tarea de modelamiento consiste en definir las características de la estructura de una red
neuronal recurrente con realimentación a la salida y luego realizar su entrenamiento usando
el algoritmo DBP (Dynamic Back-Propagation) para obtener los pesos de conexión entre las
neuronas de la red. El diseño del controlador consiste de dos etapas. En la primera etapa se
definen las características de la estructura de una red neuronal prealimentada (feed-forward).
Para el entrenamiento de la red se utiliza el algoritmo DBP bajo un enfoque dinámico donde
se considera el sistema en lazo cerrado, el cual comprende tanto al controlador como al
modelo del sistema. El controlador de posición obtenido es válido solamente dentro de un
rango de movimiento; por ello, se definen un conjunto de controladores para cada rango de
operación. En la segunda etapa, se utiliza el método difuso Takagi Sugeno para la integración
de los controladores locales y la obtención de un controlador global valido en todo el rango
de operación. El controlador obtenido se implementa y prueba mediante simulación con el
objetivo de validar su desempeño para diferentes posiciones deseadas del endoscopio.
The present study is carried out within the framework of providing assistance to the surgeon in laparoscopy, which is a surgery used to treat health problems in the abdominal area. The procedure uses a camera connected to a thin flexible tube called endoscope, which allows seeing inside the patient's abdominal area; the images obtained are used by the surgeon during the patient's treatment. An essential and correct procedure consists of moving the endoscope correctly inside the abdomen. This task is seeking to facilitate its control and open up a series of possibilities such as assisted movement, remote operation and complete automation of tasks. In order to propose a solution, currently, with advances in the field of soft robotics, soft manipulators or actuators have been designed and manufactured to be used as endoscopes, which have the ability to deform and be forced to move and reach different desired positions within its operation limits. The soft manipulator body or actuator under study has four internal chambers, which can be deformed by regulating air pressure of each chamber. In order to control and reach the desired position of endoscope final effector, where a camera will be connected, in the current work the endoscope body modeling and the position controller design are carried out as main tasks. The modeling task consists of defining the characteristics of a recurrent neural network with output feedback and then training it using the DBP (Dynamic Back-Propagation) algorithm to obtain its connection weights between network neurons. The controller design consists of two stages. In the first stage, the characteristics of a feed forward neural network are defined. For network training, the DBP algorithm is used under a dynamic approach where the closed-loop system is considered, which includes both the controller and the system model. The obtained position controller is valid only within a range of motion; therefore, a set of controllers is defined for each range of operation. In the second stage, the fuzzy Takagi Sugeno method is used to integrate the local controllers and obtain a global controller for complete endoscope operating range. The controller obtained is implemented and tested by simulation in order to validate its performance for different desired positions of endoscope final effector.
The present study is carried out within the framework of providing assistance to the surgeon in laparoscopy, which is a surgery used to treat health problems in the abdominal area. The procedure uses a camera connected to a thin flexible tube called endoscope, which allows seeing inside the patient's abdominal area; the images obtained are used by the surgeon during the patient's treatment. An essential and correct procedure consists of moving the endoscope correctly inside the abdomen. This task is seeking to facilitate its control and open up a series of possibilities such as assisted movement, remote operation and complete automation of tasks. In order to propose a solution, currently, with advances in the field of soft robotics, soft manipulators or actuators have been designed and manufactured to be used as endoscopes, which have the ability to deform and be forced to move and reach different desired positions within its operation limits. The soft manipulator body or actuator under study has four internal chambers, which can be deformed by regulating air pressure of each chamber. In order to control and reach the desired position of endoscope final effector, where a camera will be connected, in the current work the endoscope body modeling and the position controller design are carried out as main tasks. The modeling task consists of defining the characteristics of a recurrent neural network with output feedback and then training it using the DBP (Dynamic Back-Propagation) algorithm to obtain its connection weights between network neurons. The controller design consists of two stages. In the first stage, the characteristics of a feed forward neural network are defined. For network training, the DBP algorithm is used under a dynamic approach where the closed-loop system is considered, which includes both the controller and the system model. The obtained position controller is valid only within a range of motion; therefore, a set of controllers is defined for each range of operation. In the second stage, the fuzzy Takagi Sugeno method is used to integrate the local controllers and obtain a global controller for complete endoscope operating range. The controller obtained is implemented and tested by simulation in order to validate its performance for different desired positions of endoscope final effector.
Descripción
Palabras clave
Controladores programables--Diseño y construcción, Robótica--Automatización, Ingeniería biomédica--Aparatos e instrumentos
Citación
Colecciones
item.page.endorsement
item.page.review
item.page.supplemented
item.page.referenced
Licencia Creative Commons
Excepto se indique lo contrario, la licencia de este artículo se describe como info:eu-repo/semantics/openAccess