Use this link to cite:
http://hdl.handle.net/2183/39420 Implementación de aprendizaje por refuerzo en robots con patas para aprender a caminar
Loading...
Identifiers
Publication date
Authors
Rivadulla Brey, Jorge
Advisors
Other responsabilities
Universidade da Coruña. Facultade de Informática
Journal Title
Bibliographic citation
Type of academic work
Academic degree
Abstract
[Resumen]: Este trabajo de fin de grado se enmarca dentro del campo de la inteligencia artificial y de la robótica. El grupo integrado de ingeniería de la UDC (GII) ha llevado a cabo diferentes investigaciones en este campo con la intención de comprobar el funcionamiento de algoritmos evolutivos a la hora de obtener la morfología óptima para que los robots con patas aprendan a caminar. A partir de esos experimentos, y usando las mismas morfologías de los robots, nace el interés por conocer como serían los resultados usando, en este caso, aprendizaje por refuerzo. Para ello, se utiliza el simulador CoppeliaSim donde se encuentran ya implementadas diferentes morfologías entre las que destaca la del robot de cuatro patas, por su estabilidad. El objetivo final de este proyecto es el de investigar diferentes tipos de aprendizaje por refuerzo, decidir cuál es el de mayor interés e implementarlo para conseguir que dicho robot sea capaz de aprender a caminar de forma autónoma.
[Abstract]: This final degree project is part of the field of artificial intelligence and robotics. The integrated engineering group of the UDC (GII) has carried out several experiments in this field with the intention of verifying the operation of evolutionary algorithms when it comes to obtaining the optimal morphology for robots with legs to learn to walk. From these experiments, and using the same morphologies of the robots, interest was born in knowing what the results would be like using, in this case, reinforcement learning. For this, the CoppeliaSim simulator is used where different morphologies are already implemented, among which the four-legged robot stands out, due to its stability. The final objective of this project is to investigate different types of reinforcement learning, decide which is of greatest interest and implement it to ensure that said robot is capable of learning to walk autonomously.
[Abstract]: This final degree project is part of the field of artificial intelligence and robotics. The integrated engineering group of the UDC (GII) has carried out several experiments in this field with the intention of verifying the operation of evolutionary algorithms when it comes to obtaining the optimal morphology for robots with legs to learn to walk. From these experiments, and using the same morphologies of the robots, interest was born in knowing what the results would be like using, in this case, reinforcement learning. For this, the CoppeliaSim simulator is used where different morphologies are already implemented, among which the four-legged robot stands out, due to its stability. The final objective of this project is to investigate different types of reinforcement learning, decide which is of greatest interest and implement it to ensure that said robot is capable of learning to walk autonomously.
Description
Editor version
Rights
Atribución 3.0 España







