Technische Universität Darmstadt

About: Technische Universität Darmstadt is a education organization based out in Darmstadt, Germany. It is known for research contribution in the topics: Computer science & Context (language use). The organization has 17316 authors who have published 40619 publications receiving 937916 citations. The organization is also known as: Darmstadt University of Technology & University of Darmstadt.

A peer-to-peer approach to content-based publish/subscribe

Abstract: Publish/subscribe systems are successfully used to decouple distributed applications. However, their efficiency is closely tied to the topology of the underlying network, the design of which has been neglected. Peer-to-peer network topologies can offer inherently bounded delivery depth, load sharing, and self-organisation. In this paper, we present a content-based publish/subscribe system routed over a peer-to-peer topology graph. The implications of combining these approaches are explored and a particular implementation using elements from Rebeca and Chord is proven correct.

Supporting mobility in content-based publish/subscribe middleware

Abstract: Publish/subscribe (pub/sub) is considered a valuable middleware architecture that proliferates loose coupling and leverages reconfigurability and evolution. Up to now, existing pub/sub middleware was optimized for static systems where users as well as the underlying system structure were rather fixed. We study the question whether existing pub/sub middleware can be extended to support mobile and location-dependent applications. We first analyze the requirements of such applications and distinguish two orthogonal forms of mobility: the system-centric physical mobility and an application-centric logical mobility (where users are aware that they are changing location). We introduce location-dependent subscriptions as a suitable means to exploit the power of the event-based paradigm in mobile applications. Briefly spoken, location-dependency refines a subscription to accept only events related to a mobile user's current location. Implementations for both forms of mobility are presented within the content-based pub/sub middleware Rebeca, drawing from its refined routing capabilities (namely, covering and merging).

Cnidarians: An evolutionarily conserved model system for regeneration?

Abstract: Cnidarians are among the simplest metazoan animals and are well known for their remarkable regeneration capacity. They can regenerate any amputated head or foot, and when dissociated into single cells, even intact animals will regenerate from reaggregates. This extensive regeneration capacity is mediated by epithelial stem cells, and it is based on the restoration of a signaling center, i.e., an organizer. Organizers secrete growth factors that act as long-range regulators in axis formation and cell differentiation. In Hydra, Wnt and TGF-beta/Bmp signaling pathways are transcriptionally up-regulated early during head regeneration and also define the Hydra head organizer created by de novo pattern formation in aggregates. The signaling molecules identified in Cnidarian regeneration also act in early embryogenesis of higher animals. We suppose that they represent a core network of molecular interactions, which could explain at least some of the mechanisms underlying regeneration in vertebrates.

Hierarchical Relative Entropy Policy Search

Abstract: Many reinforcement learning (RL) tasks, especially in robotics, consist of multiple sub-tasks that are strongly structured. Such task structures can be exploited by incorporating hierarchical policies that consist of gating networks and sub-policies. However, this concept has only been partially explored for real world settings and complete methods, derived from first principles, are needed. Real world settings are challenging due to large and continuous state-action spaces that are prohibitive for exhaustive sampling methods. We define the problem of learning sub-policies in continuous state action spaces as finding a hierarchical policy that is composed of a high-level gating policy to select the low-level sub-policies for execution by the agent. In order to efficiently share experience with all sub-policies, also called inter-policy learning, we treat these sub-policies as latent variables which allows for distribution of the update information between the sub-policies. We present three different variants of our algorithm, designed to be suitable for a wide variety of real world robot learning tasks and evaluate our algorithms in two real robot learning scenarios as well as several simulations and comparisons.