This paper addresses the problem of path prediction for multiple interacting agents in a scene, which is a crucial step for many autonomous platforms such as self-driving cars and social robots. We present \textit{SoPhie}; an interpretable framework based on Generative Adversarial Network (GAN), which leverages two sources of information, the path history of all the agents in a scene, and the scene context information, using images of the scene. To predict a future path for an agent, both physical and social information must be leveraged. Previous work has not been successful to jointly model physical and social interactions. Our approach blends a social attention mechanism with a physical attention that helps the model to learn where to look in a large scene and extract the most salient parts of the image relevant to the path. Whereas, the social attention component aggregates information across the different agent interactions and extracts the most important trajectory information from the surrounding neighbors. SoPhie also takes advantage of GAN to generates more realistic samples and to capture the uncertain nature of the future paths by modeling its distribution. All these mechanisms enable our approach to predict socially and physically plausible paths for the agents and to achieve state-of-the-art performance on several different trajectory forecasting benchmarks.

SoPhie: An Attentive GAN for Predicting Paths Compliant to Social and Physical Constraints.

Predicting the future trajectories of multiple interacting pedestrians in a scene has become an increasingly important problem for many different applications ranging from control of autonomous vehicles and social robots to security and surveillance. This problem is compounded by the presence of social interactions between humans and their physical interactions with the scene. While the existing literature has explored some of these cues, they mainly ignored the multimodal nature of each human's future trajectory which is noticeably influenced by the intricate social interactions. In this paper, we present Social-BiGAT, a graph-based generative adversarial network that generates realistic, multimodal trajectory predictions for multiple pedestrians in a scene. Our method is based on a graph attention network (GAT) that learns feature representations that encode the social interactions between humans in the scene, and a recurrent encoder-decoder architecture that is trained adversarially to predict, based on the features, the humans' paths. We explicitly account for the multimodal nature of the prediction problem by forming a reversible transformation between each scene and its latent noise vector, as in Bicycle-GAN. We show that our framework achieves state-of-the-art performance comparing it to several baselines on existing trajectory forecasting benchmarks.

/pdf/social-bigat-multimodal-trajectory-forecasting-using-bicycle-3hfsf3wmq7.pdf

Social-BiGAT: Multimodal Trajectory Forecasting using Bicycle-GAN and Graph Attention Networks

Predicting the future trajectories of multiple interacting agents in a scene has become an increasingly important problem for many different applications ranging from control of autonomous vehicles and social robots to security and surveillance. This problem is compounded by the presence of social interactions between humans and their physical interactions with the scene. While the existing literature has explored some of these cues, they mainly ignored the multimodal nature of each human's future trajectory. In this paper, we present Social-BiGAT, a graph-based generative adversarial network that generates realistic, multimodal trajectory predictions by better modelling the social interactions of pedestrians in a scene. Our method is based on a graph attention network (GAT) that learns reliable feature representations that encode the social interactions between humans in the scene, and a recurrent encoder-decoder architecture that is trained adversarially to predict, based on the features, the humans' paths. We explicitly account for the multimodal nature of the prediction problem by forming a reversible transformation between each scene and its latent noise vector, as in Bicycle-GAN. We show that our framework achieves state-of-the-art performance comparing it to several baselines on existing trajectory forecasting benchmarks.

Pervasive computing and Internet of Things (IoTs) paradigms have created a huge potential for new business. To fully realize this potential, there is a need for a common way to abstract the heterogeneity of devices so that their functionality can be represented as a virtual computing platform. To this end, we present novel semantic level interoperability architecture for pervasive computing and IoTs. There are two main principles in the proposed architecture. First, information and capabilities of devices are represented with semantic web knowledge representation technologies and interaction with devices and the physical world is achieved by accessing and modifying their virtual representations. Second, global IoT is divided into numerous local smart spaces managed by a semantic information broker (SIB) that provides a means to monitor and update the virtual representation of the physical world. An integral part of the architecture is a resolution infrastructure that provides a means to resolve the network address of a SIB either using a physical object identifier as a pointer to information or by searching SIBs matching a specification represented with SPARQL. We present several reference implementations and applications that we have developed to evaluate the architecture in practice. The evaluation also includes performance studies that, together with the applications, demonstrate the suitability of the architecture to real-life IoT scenarios. In addition, to validate that the proposed architecture conforms to the common IoT-A architecture reference model (ARM), we map the central components of the architecture to the IoT-ARM.

Semantic Interoperability Architecture for Pervasive Computing and Internet of Things

Physical activities have tremendous benefit to older adults. A report from the World Health Organization has mentioned that lack of physical activity contributed to around 3.2 million premature deaths annually worldwide. Research also shows that regular exercise helps the older adults by improving their physical fitness, immune system, sleep and stress levels, not to mention the countless health problems it reduces such as diabetes, cardiovascular disease, dementia, obesity, joint pains, etc. The research reported in this paper is introducing a Socially Assistive Robot (SAR) that will engage, coach, assess and motivate the older adults in physical exercises that are recommended by the National Health Services (NHS) in the UK. With the rise in the population of older adults, which is expected to triple by 2050, this SAR will aim to improve the quality of life for a significant proportion of the population. To assess the proposed robot exercise trainer, user’s observational evaluation with 17 participants is conducted. Participants are generally happy with the proposed platform as a mean of encouraging them to do regular exercise correctly.

https://www.mdpi.com/2227-7080/6/1/32/pdf

Socially Assistive Robotics: Robot Exercise Trainer for Older Adults

Purpose – The purpose of this paper is to create new ideas for assistive technology products at home, especially products utilizing robotic consumer appliances available in the homes of elderly people. The work was founded on a reported increase in household robots as well as an ageing population in the industrialized world. Design/methodology/approach – Technology should be something that is perceived as belonging to our own world that fits our daily practices. Earlier studies show that in addition to cleaning functions, new household robots could change home routines and people's relationship to them. Taking the previous studies as a starting point, the paper proposes a vacuum cleaner robot as a platform for developing pervasive safety services and describe implementation of a conceptual prototype which brings the feeling of safety to an older person and their relatives by assisting in case of accidents. Moreover, the results are presented of an empirical evaluation of the prototype with end-users. Find...

An assistive household robot – doing more than just cleaning

In this work we made a preliminary clustering analysis of an experimental peer-to-peer system, tested in a small scale experiments within PlanetLab. The application was an Internet-TV like streaming system based on Chord architecture. Our clustering is inspired by the Szemeredi's Regularity Lemma (SzRL). Such approach was already demonstrated in biology and appeared to be a powerful tool. Szemeredi's result suggests that the nodes of a large enough graph can be partitioned in few clusters in such a way that link distribution between most of the pairs look like random. Our main goal is to study what can this type of clustering tell us about p2p systems using our experimental system as source of data. We searched clusterings of Szemeredi-type by using max likelihood as guidance. Our graph is directed and weighted. The link direction indicates a client-server relation and the value is the proportion of all chunks obtained from such a link during the whole experiment. We think that the preliminary results are interesting. Most of the cluster pairs have very distinguished patterns of link distribution, indicating that such a novel approach has potential in classifying peers effectively. The values of weights between clusters and their distribution show some apparent patterns. We end up with 9 cluster pairs.Contributions: practical implementations of streaming system by V. P. and analysis by H. R.

Szemerédi-type clustering of peer-to-peer streaming system

Universal Plug and Play (UPnP) is a set of specifications to enable and simplify the networking of electronic devices. UPnP does not generally provide any security and assumes that only trusted devices have access to the network. For networks where untrusted devices have to be taken into account, this paper proposes a secure UPnP network architecture, including key management. The architecture uses Transport Layer Security (TLS) to secure all TCP traffic, which carries most of UPnP messages. To establish a TLS session, each node must have an X.509 certificate for authentication. Certificates are granted by a local Certificate Authority (CA) but only if the Administrator has accepted the new node. UPnP discovery phase uses User Datagram Protocol (UDP) where it is not possible to use TLS, but we encrypt UDP data. UDP encryption key is shared by the whole network and distributed using TLS. We verified the architecture by implementation.

Secure Universal Plug and Play network

In this paper we present and evaluate a fully distributed file-sharing system architecture. Initially, a seeder node splits a large file into moderate-sized chunks and offers it for download. The seeder and all peers interested in downloading the file join a Chord-based overlay and contact each other randomly for chunk transfers. We report and discuss PlanetLab tests on this system as well as on our modified Chord implementation which it is based on. We give an algorithm for finding uniformly random peers in the overlay and another for estimating the distribution of chunk copies. Both algorithms turn out to have a significant effect on the performance, the first one in particular.

Relying on randomness PlanetLab experiments with distributed file-sharing protocols

Support of legacy devices and services is crucial for the adoption of new smart space technologies. We present two technologies which enable the formation of local ad-hoc smart spaces with commercial off-the-shelf (COTS) devices. First of these technologies is NoTA, which is a service oriented architecture enabling networks of devices with different physical transports. Second one is Smart-M3, which is a semantic information sharing architecture for smart spaces. It aims at opening physical world information for the use of services and applications in the information world, thus enabling new types of mash-up applications. In our demonstration - Open-M3 - we show how these technologies are used to build a small, yet extendable smart space for sensor monitoring using COTS devices.

Vesa Pehkonen

Papers

An assistive household robot – doing more than just cleaning

Szemerédi-type clustering of peer-to-peer streaming system

Secure Universal Plug and Play network

Relying on randomness PlanetLab experiments with distributed file-sharing protocols

Open-M3: smart space with COTS devices