MILD: Multimodal Interactive Latent Dynamics for Learning Human-Robot Interaction | IEEE Conference Publication | IEEE Xplore