BitRL-Light: 1-bit LLM Agents with Deep Reinforcement Learning for Energy-Efficient Smart Home Lighting Optimization
Ravi Gupta, Shabista Haider
TL;DR
BitRL-Light tackles the problem of energy-efficient smart home lighting by marrying 1-bit quantized LLMs with a Deep Q-Network reinforcement learning framework to run on edge devices. The system uses a three-layer architecture (voice interface, edge processing, and IoT control) and a multi-objective reward $R(s,a) = α R_{energy} + β R_{comfort} + γ R_{circadian}$ to balance energy use, user satisfaction, and circadian alignment. Key results show about $32\%$ energy savings and sub-200 ms latency on a Raspberry Pi 4, with a $5.07\times$ speedup over 2-bit models and roughly $92\%$ task accuracy, demonstrating viable cloud-free on-device AI for smart homes. This work highlights a practical path for deploying adaptive AI on affordable, resource-constrained hardware, enabling energy-aware, user-centric home automation.
Abstract
Smart home lighting systems consume 15-20% of residential energy but lack adaptive intelligence to optimize for user comfort and energy efficiency simultaneously. We present BitRL-Light, a novel framework combining 1-bit quantized Large Language Models (LLMs) with Deep Q-Network (DQN) reinforcement learning for real-time smart home lighting control on edge devices. Our approach deploys a 1-bit quantized Llama-3.2-1B model on Raspberry Pi hardware, achieving 71.4 times energy reduction compared to full-precision models while maintaining intelligent control capabilities. Through multi-objective reinforcement learning, BitRL-Light learns optimal lighting policies from user feedback, balancing energy consumption, comfort, and circadian alignment. Experimental results demonstrate 32% energy savings compared to rule-based systems, with inference latency under 200ms on Raspberry Pi 4 and 95% user satisfaction. The system processes natural language commands via Google Home/IFTTT integration and learns from implicit feedback through manual overrides. Our comparative analysis shows 1-bit models achieve 5.07 times speedup over 2-bit alternatives on ARM processors while maintaining 92% task accuracy. This work establishes a practical framework for deploying adaptive AI on resource-constrained IoT devices, enabling intelligent home automation without cloud dependencies.