Robo-Diffusion: Planning Algorithm Design

Objective

Design better planning algorithms for long-horizon trajectory generation in robot decision-making.

What You Can Modify

• Planning algorithm core logic
• Trajectory optimization strategy
• Horizon length and planning frequency
• Return conditioning method
• Trajectory selection criteria

What Is Fixed

• Network architecture (task-specific standard architecture)
• Diffusion model training
• Evaluation environments

Evaluation

Evaluated on three D4RL MuJoCo environments:

1. hopper-medium-v2
2. walker2d-medium-v2
3. halfcheetah-medium-v2

Metrics: normalized_score, episode_reward, planning_time

Baselines

diffuser

Planning with Diffusion - Original planner

decision_diffuser

Decision Diffuser - Return-conditioned planning

adaptdiffuser

AdaptDiffuser - Adaptive planning with online refinement