Catálogo de publicaciones - libros

In this paper we examine the performance advantages of using a GPU to execute the space variant Gaussian filtering. Our results show that the straightforward convolution GPU implementation obtains up to 8 times better performance than the best recursive algorithm (the Deriche’s filter) executed on a CPU, for useful maximum values. GPUs have turned out a useful option to obtain high execution performance, specially due to the emergence of high level languages for graphics hardware.

Frequent changes caused by IP-connectivity and user-oriented services in Inter-Vehicular Communication Networks (VCNs) set great challenges to construct reliable, secure and fast converged topology formed by trusted mobile nodes and links. In this paper, based on a new metric for network performance called topology convergence and a new Object-Oriented Management Information Base - active MIB (O:MIB), we propose an ant-based topology convergence algorithm that applies the swarm intelligence metaphor to find the near-optimal converged topology in VCNs which maximizes system performance and guarantee a further sustainable and maintainable system topology to achieve Quality of Service (QoS) and system throughput. This algorithm is essentially a distributed approach in that each node collects information from local neighbor nodes by invoking the methods from each localized O:MIB, through the sending and receiving of ant packets from each active node, to find the appropriate nodes to construct a routing path. Simulation results show this approach can lead to a fast converged topology with regards to multiple optimization objectives, as well as scale to network sizes and service demands.

We present a simulation based study of an arrangement consisting of a sensor array, connected to a set of signal processing units (PUs), capable to be reconfigured in real time, depending on the signal type and processing requirements. The signal issued by a sensor is conveyed by means of an arbitration algorithm to a proper PU, so that the number of its state transitions should decrease. The algorithm also ensures a high throughput of the signals to be processed. The system includes queues at the PU level, to lower data loss. The functioning of the PUs is based upon two strategies: maximising task charge and standby state period, to maintain the highest number of spare PUs in order to decrease the reconfiguration actions and the associate power penalty. The sensors are capable to issue signals with variable frequency in time, a factor considered when prioritising among concomitantly generated signals.

Robot soccer was introduced with the purpose to develop intelligent cooperative multi-robot (agents) systems (MAS). From the scientific point of view a soccer robot is an intelligent, autonomous agent, carrying out tasks together with other agents in a cooperative, coordinated and communicative way. Robot soccer is a good test bed for the development of MAS for current and future industrial applications. The robots in a team have a common goal – to kick the ball in the opponent goal and to avoid goals against the own team. The cooperation and coordination of actions by means of communication are necessary. In this contribution the development of three generations of mini robots for robot soccer, in the FIRA categories Miro- and Narosot, are described. Finally “Roby Space” as a “spin-off” of robot soccer is presented.

We present an embedded vision sensor to be used for robot soccer in the MiroSot league of the Federation of the International Robotsoccer Association. The vision sensor is based on a DSP/FPGA co-processor system and uses a FireWire-camera. The vision algorithms are specifically designed to optimally utilize the resources of the embedded system. The achieved embedded vision sensor is able to work at full camera framerate (60fps) with full image resolution (640x480 pixels) without the need of any resources of the host computer.

Sociable robots are gaining popularity among robotics researchers and cognitive scientists. These robots generally show abilities that include expressive power (face, voice, ...), locating, paying attention to and addressing people, etc. Such abilities fall into what is known as social intelligence in humans. The reproduction of social intelligence, as opposed to other types of human abilities, may lead to fragile performance, in the sense of having rather different performances between tested cases and future (untested) cases and situations. This limitation stems from the fact that our social abilities are mainly unconscious to us. This is in contrast with other human abilities that we carry out using conscious effort, and for which we can easily conceive algorithms and representations. The fragile performance mentioned above is nothing but overfitting. Thus, we propose to approach the problem using strategies followed in Machine Learning for avoiding overfitting. Specifically, complexity penalization and incremental design are translated to the broader ambit of robot design and development. The robot CASIMIRO is currently being developed following that approach.

This paper presents the ”robuCAB”, a Cybercar developed by ROBOSOFT, and the results of experimentation in a public and pedestrian area. In a first part, the hardware and software architecture of the vehicle - control, safety and HMI - are presented. The results of experimentation are detailed and the perspectives of evolution (control, safety, HMI) are discussed.

This paper address the perception of the driving environment of an Autonomous Guided Vehicle from a vision system. The trajectory planning for an Autonomous Guided Vehicle requires to robustly segment the free space around the vehicle to manage an obstacle avoidance task. The urban environment still remains as the most complex environment due to the variety of dynamic elements which animate the scenes. The feature-based approach we detail in this article consists first on the segmentation of the ground surface to reconstruct finally a 2D polar map of the obstacles edges.

Several single classifiers have been proposed to recognize objects in images. Since this approach has restrictions when applied in certain situations, one has suggested some methods to combine the outcomes of classifiers in order to increase overall classification accuracy. In this sense, we propose an effective method for a frame-by-frame classification task, in order to obtain a trade-off between false alarm decrease and true positive detection rate increase. The strategy relies on the use of a Class Set Reduction method, using a Mamdani fuzzy system, and it is applied to recognize pedestrians and vehicles in typical cybercar scenarios. The proposed system brings twofold contributions: i) overperformance with respect to the component classifiers and ii) expansibility to include other types of classifiers and object classes. The final results have shown the effectiveness of the system.

Even though computer-based navigation systems, being developed since the late 1980s, are still among the key topics worked on in information technology and the road network changes circa 15% per year, there are no significant possibilities for update and adjustment of the digital map data.

In this paper, we present the concepts of our software system under development, which aims at the automatic construction and extension of digital maps, thus enabling continuous improvement of their quality as well as allowing for advanced, intelligent, and personalizable navigation, depending on user behavior and preference. This approach comprises an appropriate processing of the raw data, reliable graph optimization and merging techniques, and finally, suitable data exchange interfaces.