In this paper, a grounding framework is proposed that combines unsupervised and supervised grounding by extending an unsupervised grounding model with a mechanism to learn from explicit human teaching. To investigate whether explicit teaching improves the sample efficiency of the original model, both models are evaluated through an interaction experiment between a human tutor and a robot in which synonymous shape, color, and action words are grounded through geometric object characteristics, color histograms, and kinematic joint features. The results show that explicit teaching improves the sample efficiency of the unsupervised baseline model.

No abstract

We investigate how reliable movement can emerge in aggregates of highly error-prone individuals. The individuals—robotic modules—move stochastically using vibration motors. By coupling them via elastic links, soft-bodied aggregates can be created. We present distributed algorithms that enable the aggregates to move and deform reliably. The concept and algorithms are validated through formal analysis of the elastic couplings and experiments with aggregates comprising up to 49 physical modules—among the biggest soft-bodied aggregates to date made of autonomous modules. The experiments show that aggregates with elastic couplings can shrink and stretch their bodies, move with a precision that increases with the number of modules, and outperform aggregates with no, or rigid, couplings. Our findings demonstrate that mechanical couplings can play a vital role in reaching coherent motion among individuals with exceedingly limited and error-prone abilities, and may pave the way for low-power, stretchable robots for high-resolution monitoring and manipulation.

The allocation of tasks to members of a team is a well-studied problem in robotics. Applying market-based mechanisms, particularly auctions, is a popular solution. We focus on evaluating the performance of the team when executing the tasks that have been allocated. The work presented here examines the impact of one such factor, namely task duration. Building on prior work, a new bidding strategy and performance metric are introduced. Experimental results are presented showing that there are statistically significant differences in both time and distance-based performance metrics when tasks have zero vs greater-than-zero duration.

This paper presents the results of preliminary experiments in humanrobot collaboration for an agricultural task. CCS CONCEPTS• Human-centered computing → Human computer interaction (HCI); • Computer systems organization → Robotic control; • Computing methodologies → Vision for robotics.