Efficient Diffusion Planning with Temporal Diffusion

Diffusion planning is a promising method for learning high-performance policies from offline data. To avoid the impact of discrepancies between planning and reality on performance, previous works generate new plans at each time step. However, this incurs significant computational overhead and leads to lower decision frequencies, and frequent plan switching may also affect performance. In contrast, humans might create detailed short-term plans and more general, sometimes vague, long-term plans, and adjust them over time. Inspired by this, we propose the Temporal Diffusion Planner (TDP) which improves decision efficiency by distributing the denoising steps across the time dimension. TDP begins by generating an initial plan that becomes progressively more vague over time. At each subsequent time step, rather than generating an entirely new plan, TDP updates the previous one with a small number of denoising steps. This reduces the average number of denoising steps, improving decision efficiency. Additionally, we introduce an automated replanning mechanism to prevent significant deviations between the plan and reality. Experiments on D4RL show that, compared to previous works that generate new plans every time step, TDP significantly improves the decision-making frequency by 11-24.8 times while achieving higher or comparable performance.

Authors

• Jiaming Guo State Key Lab of Processors, Institute of Computing Technology, CAS
• Rui Zhang
• Zerun Li SKL of Processors, Institute of Computing Technology, CAS University of Chinese Academy of Sciences Cambricon Technologies
• Yunkai Gao University of Science and Technology of China, USTC, Hefei, China
• Shaohui Peng Intelligent Software Research Center, Institute of Software, CAS
• Siming Lan University of Science and Technology of China, USTC, Hefei, China
• Xing Hu SKL of Processors, Institute of Computing Technology, CAS
• Zidong Du SKL of Processors, Institute of Computing Technology, CAS
• Xishan Zhang SKL of Processors, Institute of Computing Technology, CAS
• Ling Li

Keywords

No keywords are indexed for this paper.