A concise overview of different reinforcement learning algorithms for robotics exploration

The ambition to enhance the intelligence of robots and computers, along with the goal of facilitating their autonomous operation, has propelled advancements in neural networks, deep learning, and various artificial intelligence methodologies. While reinforcement learning has predominantly been appli...

Descripción completa

Detalles Bibliográficos
Autores: Raj, Ravi|||0000-0001-8073-1812, Kos, Andrzej
Tipo de recurso: artículo
Fecha de publicación:2025
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/450508
Acceso en línea:https://hdl.handle.net/2117/450508
https://dx.doi.org/10.15199/48.2025.09.35
Access Level:acceso abierto
Palabra clave:Artificial intelligence (AI)
Autonomous robot
Deep learning (DL)
Path planning
Reinforcement learning (RL)
Àrees temàtiques de la UPC::Informàtica::Intel·ligència artificial
Descripción
Sumario:The ambition to enhance the intelligence of robots and computers, along with the goal of facilitating their autonomous operation, has propelled advancements in neural networks, deep learning, and various artificial intelligence methodologies. While reinforcement learning has predominantly been applied in video games, recent progress and the emergence of various robust reinforcement algorithms have facilitated the transition of the reinforcement learning group from gaming to addressing intricate real-world challenges in autonomous systems, including self-driving vehicles, transportation drones, and autonomous robots. Comprehending the application's surroundings and the constraints of algorithms is important for determining the suitable reinforcement learning algorithm that effectively addresses the issue at hand. Thus, this study provides the basic information of reinforcement learning based algorithms that are useful, especially in the context of robotics perceptions, path planning, and obstacle avoidance. Furthermore, within each group, we discern correlations among algorithms. The general description of each algorithm explains its foundational principles and examines the relationships and variances among them. This study offers insights into the algorithms and assists professionals and academics in selecting the suitable algorithm for its particular use scenario.