Efficient Diffusion Planning with Temporal Diffusion
This work addresses a computational bottleneck in diffusion planning for offline RL, offering a more efficient method for real-time decision-making.
The paper tackles the computational inefficiency of diffusion planning in offline reinforcement learning by proposing the Temporal Diffusion Planner (TDP), which distributes denoising steps across time to reduce overhead, resulting in an 11-24.8 times improvement in decision-making frequency while maintaining or enhancing performance on D4RL benchmarks.
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 improves the decision-making frequency by 11-24.8 times while achieving higher or comparable performance.