The paper introduces DELTA, a novel approach for efficient long-term robot task planning using Large Language Models (LLMs). DELTA integrates scene graphs (SGs) and LLMs to generate precise planning problem descriptions and decompose long-term goals into sub-goals, enhancing planning performance and computational efficiency. The system architecture consists of five steps: domain generation, scene graph pruning, problem generation, goal decomposition, and autoregressive sub-task planning. DELTA outperforms existing LLM-based baselines in terms of success rates, plan quality, and planning time, demonstrating its effectiveness in solving complex long-term tasks in large and complex environments. The evaluation covers five domains, including laundry, PC assembly, dining table setup, house cleaning, and home office setup, using various metrics such as success rate, plan length, planning time, and number of expanded nodes. The results show that DELTA significantly improves planning efficiency and success rates compared to other approaches.The paper introduces DELTA, a novel approach for efficient long-term robot task planning using Large Language Models (LLMs). DELTA integrates scene graphs (SGs) and LLMs to generate precise planning problem descriptions and decompose long-term goals into sub-goals, enhancing planning performance and computational efficiency. The system architecture consists of five steps: domain generation, scene graph pruning, problem generation, goal decomposition, and autoregressive sub-task planning. DELTA outperforms existing LLM-based baselines in terms of success rates, plan quality, and planning time, demonstrating its effectiveness in solving complex long-term tasks in large and complex environments. The evaluation covers five domains, including laundry, PC assembly, dining table setup, house cleaning, and home office setup, using various metrics such as success rate, plan length, planning time, and number of expanded nodes. The results show that DELTA significantly improves planning efficiency and success rates compared to other approaches.