Robots are well-suited to alleviate the burden of repetitive and tedious manipulation tasks. In many applications, though, a robot may be asked to interact with a wide variety of objects, making it hard or even impossible to pre-program visual object classifiers suitable for the task of interest. In this work, we study the problem of learning a classifier for visual objects based on a few examples provided by humans. We frame this problem from the perspective of learning a suitable visual object representation that allows us to distinguish the desired object category from others. Our proposed approach integrates human supervision into the representation learning process by combining contrastive learning with an additional loss function that brings the representations of human examples close to each other in the latent space. Our experiments show that our proposed method performs better than self-supervised and fully supervised learning methods in offline evaluations and can also be used in real-time by a robot in a simplified recycling domain, where recycling streams contain a variety of objects.

We use over 500 unique recyclable items of various sizes, shapes, colors, and forms in our experiments. The recyclables belong to ten categories: crushed metal cans, un-crushed metal cans, crushed plastic bottles, un-crushed plastic bottles, un-crushed colored plastic bottles, brown cardboard boxes, coated cardboard boxes, cardboard trays, half-gallon milk jugs, and one-gallon milk jugs. For randomizing the stream, the hopper dispenses a random selection of recyclables onto the feeder belt, which then transports the items to the main conveyor belt. A human can pick up some items of a given category by standing next to the conveyor belt. The robot is expected to sort items similar to the human-selected objects and extract them from the stream by pushing or pulling them off the conveyor belt.

@inproceedings{ghose2022tailoring,

title={Tailoring Visual Object Representations to Human Requirements: A Case Study with a Recycling Robot},

author={Ghose, Debasmita and Lewkowicz, Michal Adam and Gezahegn, Kaleb and Lee, Julian and Adamson, Timothy and Vazquez, Marynel and Scassellati, Brian},

booktitle={6th Annual Conference on Robot Learning},

year={2022}

}