"Cognitive Agentic AI Framework for Smart Home Automation with Reinforcement Learning and Trust-Aware Decision Making "

"The emergence of more advanced smart home systems has created an ever-increasing demand of fully automatic, autonomous and intelligent service management systems. However, the current solutions do not guarantee unified decision-making across services and do not warrant the inherent trust and real-time adaptation-consciousness that smart homeowners require. In this paper, the authors introduce a multi-agents system architecture, CAREHome: Cognitive Agentic Reinforcement-enhanced Home Automation Framework, which integrates Large Language Models (LLMs), Reinforcement Learning (RL) algorithms and context-adaptive trust-based decision-making dynamics into smart homes setups to coordinate the delivery of provided services. It synthesize structured intents automatically based on user inputs, dynamically evaluate the performance of service providers based on trust values and optimize actions based on reinforcement learning strategies. An exhaustive experimental validation over 1000 service interactions synthetic dataset in simulated scenario is provided: 96.3% of goal was correctly understood; success rate reached 93.8%, considering booking processes; consumer waits 1.52 seconds less averagely when compares all requests meanwhile having their satisfaction increased with the proposed approach relatively to baselines even when they are reduced too. Results show that our contribution is the first step that will actually ensure such consolidated automatic adaptive decisions that meet the demanded needs set by developing homes automation."