HeteroSense-FL: A multimodal simulation testbed for modality-heterogeneous federated learning
Xun Shao, Kohsuke Yamakawa, Otani Aoba · SoftwareX · 2026
HeteroSense-FL is a Python software package for structured multimodal sensor simulation targeting modality-heterogeneous federated learning (FL) research. It generates 3D LiDAR point clouds and bed pressure maps across N client sites, each configured with a different sensor subset, enabling systematic study of FL algorithms under per-client modality heterogeneity. The package provides a configurable N -client heterogeneity API, a temporal window sampler for plug-and-play encoder development, and automated observation integrity checks. A reference benchmark is reproduced with the single command heterosense-benchmark . The software is CI-tested across Python 3.9 to 3.12 on Linux, macOS, and Windows.