Interpretable Latent Spaces for Learning from Demonstration

Abstract

Effective human-robot interaction, such as in robot learning from human demonstration, requires the learning agent to be able to ground abstract concepts (such as those contained within instructions) in a corresponding high-dimensional sensory input stream from the world. Models such as deep neural networks, with high capacity through their large parameter spaces, can be used to compress the high-dimensional sensory data to lower dimensional representations. These low-dimensional representations facilitate symbol grounding, but may not guarantee that the representation would be human-interpretable. We propose a method which utilises the grouping of user-defined symbols and their corresponding sensory observations in order to align the learnt compressed latent representation with the semantic notions contained in the abstract labels. We demonstrate this through experiments with both simulated and real-world object data, showing that such alignment can be achieved in a process of physical symbol grounding.

arXiv comment12 pages, 6 figures, accepted at the Conference on Robot Learning (CoRL) 2018, Zurich, Switzerland

Cited by in corpus (10)

• COG: Connecting New Skills to Past Experience with Offline Reinforcement Learning
• embComp: Visual Interactive Comparison of Vector Embeddings
• A Workflow for Offline Model-Free Robotic Reinforcement Learning
• Vid2Param: Modelling of Dynamics Parameters from Video
• Preventing Imitation Learning with Adversarial Policy Ensembles
• Disentangled Relational Representations for Explaining and Learning from Demonstration
• Learning from Demonstration with Weakly Supervised Disentanglement
• Learning Interpretable Concept-Based Models with Human Feedback
• Neural Embedding for Physical Manipulations
• Generation Drawing/Grinding Trajectoy Based on Hierarchical CVAE