Discovering the linguistic structure of a language solely from spoken input asks for two steps: phonetic and lexical discovery. The first is concerned with identifying categorical subword unit inventory relating it to underlying acoustics, while second aims at discovering words as repeated patterns units. hierarchical approach presented here accounts classification errors in stage by modelling pronunciation word terms units probabilistically: hidden Markov model discrete emission...

Locomotion is a prime example for adaptive behavior in animals and biological control principles have inspired architectures legged robots. While machine learning has been successfully applied to many tasks recent years, Deep Reinforcement Learning approaches still appear struggle when real world robots continuous particular do not as robust solutions that can handle uncertainties well. Therefore, there new interest incorporating into such architectures. inducing hierarchical organization...

Deep Reinforcement Learning techniques demonstrate advances in the domain of robotics. One limiting factors is a large number interaction samples usually required for training simulated and real-world environments. In this work, we set dexterous in-hand object manipulation tasks that tactile information can substantially increase sample efficiency (by up to more than threefold). We also observe an improvement performance (up 46%) after adding information. To examine role tactile-sensor...

AMiRo is a novel modular robot platform that can be easily extended and customized in hardware software. Built up of electronic modules fully exploit recent technology open-source software for sensor processing, actuator control, cognitive the facilitates rich autonomous behaviors. Further contribution lies completely habitat: from low-level microcontroller implementations, over high-level applications running on an embedded processor, to accelerated algorithms using programmable logic. This...

In the past, several contributions and proposals for implementation of Ultra-wideband (UWB)-based localization positioning solutions on system level were made. However, most them are limited to a unidirectional approach, i.e. data communication is in one direction (from transmitter receiver). This restricts systems' use-case either navigation or tracking. this paper, we demonstrate an UWB-based bidirectional which capable acting as both tracking single wireless platform. Regarding this,...

This work presents the novel multi-modal Variational Autoencoder approach <tex>$\mathbf{M}^{\mathbf{2}}\mathbf{VAE}$</tex> which is derived from complete marginal joint log-likelihood. allows end-to-end training of Bayesian information fusion on raw data for all subsets a sensor setup. Furthermore, we introduce concept in-place &#x2013; applicable to distributed sensing - where latent embeddings observations need be fused with new data. To facilitate even data, introduced re-encoding loss...

Today, agricultural vehicles are available that can automatically perform tasks such as weed detection and spraying, mowing, sowing while being steered automatically. However, for systems to be fully autonomous self-driven, not only their specific must automated. An accurate robust perception system detecting avoiding all obstacles also realized ensure safety of humans, animals, other surroundings. In this paper, we present a multi-modal obstacle environment recognition approach process...

In this paper, we analyze five true-range positioning algorithms for UWB-based localization systems.The evaluated are: (i) trilateration using a geometric method, (ii) closed-form multilateration solution least squares, (iii) an iterative approach first-order Taylor series, recursive based on (iv) the Extended Kalman Filter (EKF), and (v) Unscented (UKF).In contrast to existing comparative studies in literature, which are solely simulation results, our analysis is experimental...

In recent decades, mapping has been increasingly investigated and applied in unmanned terrain, aerial, sea, underwater vehicles. While exploiting various techniques to build an inner representation of the environment, one most famous remaining is occupancy grid mapping. It all domains a 2D/3D fashion for localization, mapping, navigation, safe path traversal. Until now generally active range measuring sensors like LiDAR or SONAR are exploited those maps. With this work authors want overcome...

We present a novel approach of multi-modal deep generative models and apply this to coordinated heterogeneous multi-agent active sensing. A major achieve objective is train variational Auto Encoder (M <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> VAE) that integrates the information different sensor modalities into joint latent representation. Furthermore, we derive an from M VAE enables maximization evidence lower bound via selection...

A huge number of techniques for detecting and mapping obstacles based on LIDAR SONAR exist, though not taking approximative sensors with high levels uncertainty into consideration. The proposed mapping method in this article is undertaken by surfaces approximating objects distance using sensors with localization ambiguity. Detection an Inverse Particle Filter, which uses readings from single or multiple as well a robot’s motion. This contribution describes the extension...

Evaluation of robotic experiments requires physical robots as well position sensing systems. Accurate systems detecting sufficiently all necessary degrees freedom, like the famous Vicon system, are commonly too expensive. Therefore, we target an economical multi-camera based solution by following these three requirements: Using multiple cameras to track even large laboratory areas, applying fiducial marker trackers for pose identification, and fuse tracking hypothesis resulting from via...

Deep Reinforcement Learning (DRL) approaches have shown tremendous success over the last years in different application areas. But control of robots real world settings and when facing unpredictable environments has still proven to be a difficult task that requires unreasonable long training times. This sparked new interest organization animal human motor systems how transfer these insights into such DRL learning architectures. While hierarchical now been advocated introduced couple...

In recent years, the drive of Industry 4.0 initiative has enriched industrial and scientific approaches to build self-driving cars or smart factories. Agricultural applications benefit from both advances, as they are in reality mobile driving factories which process environment. Therefore, acurate perception surrounding is a crucial task it involves goods be processed, contrast standard indoor production lines. Environmental processing requires accurate robust quantification order correctly...

We investigate a reinforcement approach for distributed sensing based on the latent space derived from multi-modal deep generative models. Our contribution provides insights to following benefits: Detections can be exchanged effectively between robots equipped with uni-modal sensors due shared representation of information that is trained by Variational Auto Encoder (VAE). Sensor-fusion applied asynchronously feature VAE. Deep Q-Networks (DQNs) are minimize uncertainty in coordinating...

With the continuous progress in robotics and application of such systems evermore scenarios, safety flexibility become increasingly important aspects new designs should thus emphasize real-time capability modularity. This work points out all related topics for an endeavor proclaims to move from conventional bottom-up design more holistic approaches. Based on experience gained with modular mini robot platforms BeBot AMiRo, a novel generic architecture is proposed that offers high system wide...