A survey on 3D Perception for robot object manipulation tasks
Fecha
2019
Tipo
comunicación de congreso
Autores
Peralta Sáenz, Javier Ignacio
Ruiz Ugalde, Federico
Título de la revista
ISSN de la revista
Título del volumen
Editor
Resumen
Un sistema confiable de manipulación de objetos requiere de percepción sofisticada.
Aspectos importantes de este sistema es ser capaz de detectar la forma, tamaño y pose de los objetos a
manipular. Estas capacidades se han mostrado antes en la literatura pero sin demostrar altos grados de
confiabilidad en ambientes no controlados. Las soluciones actuales tienen muchas limitaciones para
lograr un manejo adecuado de los objetos. El objetivo principal de este estudio es revisar las técnicas
de percepción más relevantes al problema, sus características y limitaciones. Esto será de utilidad para
dirigir el trabajo futuro a los aspectos que un robot (como el desarrollado en el ARCOS-Lab) necesita
para manipular objetos en ambientes humanos no controlados cotidianos, como las cocinas
A reliable object manipulation system requires a sophisticated perception. Important aspects that are required for this type of manipulation are been able to detect the shape, size and pose of the objects to manipulate. These capabilities have been shown before but their reliability is usually not high enough to achieve proper object manipulation in uncontrolled environments. Many are the limitations of current solutions and many are the requirements for this type of manipulation. The main goal of this survey is to study the current most relevant methods, their features and limitations. This will be useful to concentrate future work on the missing aspects that our robots will need to achieve a reliable object manipulation in unconstrained every day human environments such as kitchens.
A reliable object manipulation system requires a sophisticated perception. Important aspects that are required for this type of manipulation are been able to detect the shape, size and pose of the objects to manipulate. These capabilities have been shown before but their reliability is usually not high enough to achieve proper object manipulation in uncontrolled environments. Many are the limitations of current solutions and many are the requirements for this type of manipulation. The main goal of this survey is to study the current most relevant methods, their features and limitations. This will be useful to concentrate future work on the missing aspects that our robots will need to achieve a reliable object manipulation in unconstrained every day human environments such as kitchens.
Descripción
Palabras clave
Robótica, Manipulación de objetos, Percepción, Segmentación semántica, Estimación de pose, Estimación de parámetros, Robotics, Object manipulation, Perception, Semantic segmentation, Pose estimation, Parameter estimation