Sparse Federated Representation Learning for bio-inspired soft robotics maintenance under real-time policy constraints

Short summary

A federated learning framework for maintaining soft robots across distributed facilities while respecting data privacy and real-time constraints. Uses sparse autoencoders to learn compressed representations enabling sub-10ms inference on edge hardware without centralizing proprietary manufacturing data.

• •Federated learning solves data privacy and network bandwidth constraints in soft robotics maintenance
• •Sparse representation learning achieves sub-10ms real-time inference on edge devices
• •Framework handles heterogeneous sensors across multiple facilities without data centralization

Generated with AI, which can make mistakes.