Sparse Compositional Flow Matching by geometric assembly from motion primitives

1. 1Implement Motion-Primitive Dictionary Learning to equip atoms with learnable length masks and binary starting indicators
2. 2Apply Structural Sparse Flow Matching with Geometric Constraints to generate a binary placement matrix
3. 3Use duration-aware tokenization and differentiable geometric loss to enforce spatial continuity and temporal contiguity
4. 4Evaluate the framework on Open X-Embodiment and 3DMoTraj datasets to assess its accuracy and efficiency
5. 5Fine-tune the framework to improve ADE and FDE ratios

Robotics and AI engineers can benefit from this framework to improve the efficiency and accuracy of embodied trajectory generation, enabling better task decomposition and motion planning

💡 Compositional latent structure can be used to generate embodied trajectories, improving accuracy and reducing errors

💡 Improve embodied trajectory generation with sparse compositional flow matching!