We present two empirical studies on the design of control software for robot swarms. In Study A, Vanilla and EvoStick, two previously published automatic design methods, are compared with human designers. The comparison is performed on five swarm robotics tasks that are different from those on which Vanilla and EvoStick have been previously tested. The results show that, under the experimental conditions considered, Vanilla performs better than EvoStick, but it is not able to outperform human designers. The results indicate that Vanilla's weak element is the optimization algorithm employed The main contributors to this research are G. Francesca and M. Birattari. AutoMoDe and Vanilla were conceived and developed by G. Francesca, M. Brambilla, A. Brutschy, V. Trianni, and M. Birattari.123 Swarm Intell to search the space of candidate designs. To improve over Vanilla and with the final goal of obtaining an automatic design method that performs better than human designers, we introduce Chocolate, which differs from Vanilla only in the fact that it adopts a more powerful optimization algorithm. In Study B, we perform an assessment of Chocolate. The results show that, under the experimental conditions considered, Chocolate outperforms both Vanilla and the human designers. Chocolate is the first automatic design method for robot swarms that, at least under specific experimental conditions, is shown to outperform a human designer.

The performance of deep neural networks is strongly influenced by the quantity and quality of annotated data. Most of the large activity recognition datasets consist of data sourced from the web, which does not reflect challenges that exist in activities of daily living. In this paper, we introduce a large real-world video dataset for activities of daily living: Toyota Smarthome. The dataset consists of 16K RGB+D clips of 31 activity classes, performed by seniors in a smarthome. Unlike previous datasets, videos were fully unscripted. As a result, the dataset poses several challenges: high intra-class variation, high class imbalance, simple and composite activities, and activities with similar motion and variable duration. Activities were annotated with both coarse and fine-grained labels. These characteristics differentiate Toyota Smarthome from other datasets for activity recognition. As recent activity recognition approaches fail to address the challenges posed by Toyota Smarthome, we present a novel activity recognition method with attention mechanism. We propose a pose driven spatiotemporal attention mechanism through 3D ConvNets. We show that our novel method outperforms state-of-the-art methods on benchmark datasets, as well as on the Toyota Smarthome dataset. We release the dataset for research use 1 .

Designing collective behaviors for robot swarms is a difficult endeavor due to their fully distributed, highly redundant, and ever-changing nature. To overcome the challenge, a few approaches have been proposed, which can be classified as manual, semi-automatic, or automatic design. This paper is intended to be the manifesto of the automatic off-line design for robot swarms. We define the off-line design problem and illustrate it via a possible practical realization, highlight the core research questions, raise a number of issues regarding the existing literature that is relevant to the automatic off-line design, and provide guidelines that we deem necessary for a healthy development of the domain and for ensuring its relevance to potential real-world applications.

Robot swarms mimic natural systems in which collective abilities emerge from the interaction of individuals. So far, the swarm robotics literature has focused on the emergence of mechanical abilities (e.g., push a heavy object) and simple cognitive abilities (e.g., select a path between two alternatives). In this article, we present a robot swarm in which a complex cognitive ability emerged. This swarm was able to collectively sequence tasks whose order of execution was a priori unknown. Because sequencing tasks is an albeit simple form of planning, the robot swarm that we present provides a different perspective on a pivotal debate in the history of artificial intelligence: the debate on planning in robotics. In the proposed swarm, the two robotics paradigms-deliberative (sense-modelplan-act) and reactive (sense-act)-that are traditionally considered antithetical coexist in a particular way: The ability to plan emerges at the collective level from the interaction of reactive individuals.

Handling long and complex temporal information is an important challenge for action detection tasks. This challenge is further aggravated by densely distributed actions in untrimmed videos. Previous action detection methods fail in selecting the key temporal information in long videos. To this end, we introduce the Dilated Attention Layer (DAL). Compared to previous temporal convolution layer, DAL allocates attentional weights to local frames in the kernel, which enables it to learn better local representation across time. Furthermore, we introduce Pyramid Dilated Attention Network (PDAN) which is built upon DAL. With the help of multiple DALs with different dilation rates, PDAN can model short-term and long-term temporal relations simultaneously by focusing on local segments at the level of low and high temporal receptive fields. This property enables PDAN to handle complex temporal relations between different action instances in long untrimmed videos. To corroborate the effectiveness and robustness of our method, we evaluate it on three densely annotated, multi-label datasets: Mul-tiTHUMOS, Charades and Toyota Smarthome Untrimmed (TSU) dataset. PDAN is able to outperform previous stateof-the-art methods on all these datasets."Time abides long enough for those who make use of it." Leonardo da Vinci

We present a novel technology that allows real robots to perceive an augmented reality environment through virtual sensors. Virtual sensors are a useful and desirable technology for research activities because they allow researchers to quickly and efficiently perform experiments that would otherwise be more expensive, or even impossible. In particular, augmented reality is useful (i) for prototyping and assessing the impact of new sensors before they are physically produced; and (ii) for developing and studying the behaviour of robots that should deal with phenomena that cannot be easily reproduced in a laboratory environment because, for example, they are dangerous (e.g., fire, radiations). We realised an augmented reality system for robots in which a simulator retrieves real-time data on the real environment through a multi-camera tracking system and delivers post-processed information to the robot swarm according to each robot's sensing range. We illustrate the proposed virtual sensing technology through an experiment involving 15 e-pucks.

Abstract. We present an experiment in automatic design of robot swarms. For the first time in the swarm robotics literature, we perform an objective comparison of multiple design methods: we compare swarms designed by two automatic methods-AutoMoDe-Vanilla and EvoStickwith swarms manually designed by human experts. AutoMoDe-Vanilla and EvoStick have been previously published and tested on two tasks. To evaluate their generality, in this paper we test them without any modification on five new tasks. Besides confirming that AutoMoDe-Vanilla is e↵ective, our results provide new insight into the design of robot swarms. In particular, our results indicate that, at least under the adopted experimental protocol, not only does automatic design su↵er from the reality gap, but also manual design. The results also show that both manual and automatic methods benefit from bias injection. In this work, bias injection consists in restricting the design search space to the combinations of pre-existing modules. The results indicate that bias injection helps to overcome the reality gap, yielding better performing robot swarms.

No abstract