Diseño e implementación de un sistema basado en aprendizaje automático que facilite la percepción robótica del entorno por medio de sensores láser
The project presented below, has as an objective to realize a system based on machine learning methods, that allows a mobile robot to recognize and identify the objects found in the internal environments of the Autónoma de Occidente University, by using a low cost LIDAR sensor. Because at the presen...
- Autores:
-
Llanos Neuta, Nicolas
Aponte Vivas, Sebastian
- Tipo de recurso:
- Trabajo de grado de pregrado
- Fecha de publicación:
- 2018
- Institución:
- Universidad Autónoma de Occidente
- Repositorio:
- RED: Repositorio Educativo Digital UAO
- Idioma:
- spa
- OAI Identifier:
- oai:red.uao.edu.co:10614/10526
- Acceso en línea:
- http://hdl.handle.net/10614/10526
- Palabra clave:
- Ingeniería Mecatrónica
Aprendizaje automático (Inteligencia artificial)
Sensores remotos
Robótica
Point Cloud Library
ROS
Reconocimiento de objetos 3D
Machine learning
- Rights
- openAccess
- License
- Derechos Reservados - Universidad Autónoma de Occidente
| Summary: | The project presented below, has as an objective to realize a system based on machine learning methods, that allows a mobile robot to recognize and identify the objects found in the internal environments of the Autónoma de Occidente University, by using a low cost LIDAR sensor. Because at the present time there is no sensor that gives a robot the capability to perceive and recognize the environment in which it is located. For this purpose, a system based on free software was designed, where the data acquisition was carried out together with one of the groups belonging to the robotics research group of the UAO directed by Dr. Victor Adolfo Romero Cano, conformed by Natali Velandia and Otoniel Rodríguez, making use of the laser sensor "hokuyo urg-04lx-ug01". In addition, different methods for processing point clouds were studied, since these are the final result of data acquisition with a laser sensor, the different methods for extracting characteristics were also studied and finally different machine learning techniques such as multi-layer perceptron and support vector machines were investigated and implemented to perform the classification. The final result was an integrated system, where the expectations were met, since the delicate processing of the point clouds and the correct extraction of characteristics contributed to the correct performance of the implemented classification methods |
|---|