12/05/2026
🤖 How do autonomous software agents actually learn?
Our latest review analyzes the AI methods driving autonomous behavior, from neural networks and reinforcement learning to emerging LLM-based agents.
🔗 Read the full article here: peerj.com/articles/cs-3756/
🤖 Reinforcement Learning dominates autonomous agent learning.
🧠 Neural networks drive perception and decision-making in autonomous agents.
📚 42 studies analyzed from an initial pool of 569 records.
🔄 RNNs and CNNs were among the most used architectures.
🧪 Most evaluations were conducted in simulation environments.
🚀 LLMs are expanding the capabilities of autonomous agents.
⚡ The review identifies three learning types: aspect-based, preventive, and progressive.
🌍 Preventive learning focuses on safer agent behavior.
🎯 Progressive learning enables agents to pursue multiple objectives.
👏 Congratulations to Jorge Hernandez and the whole team for this great contribution representing SDAS Research Group.
🔍 Discover more of our work:
https://www.sdas-group.com/ourpublications/
