This paper introduces two applications of Discrete Time Cellular Non-Linear Networks (DTCNN) in a robot guiding avoiding obstacles algorithm and prove the feasibility of both applications: a high ...data rate one, using a CMOS camera, and small data rate one, using ultrasonic sensors. The key value of DTCNNs is the locally connections and the parallelism in processing. These characteristics permit a hardware implementation, in our case over a Field Programmable Gate Arraw (FPGA) and a real time template based algorithm processing. A camera and an ultrasonic sensor are used as avoiding obstacles system, requiring both implementations, different inputs informations: the first one complex environment information and the later for basic situations information where impulsive response is required. Both input can have an enhanced behaviour within DTCNN structure.
Ring Imaging Cherenkov detectors (RICH) are a class of particle detectors used for particle identification whose principle involves a pattern recognition problem: identifying circles from a short ...number of their points. A real image hosts several circles which may eventually overlap. Actual techniques solving this problem are mainly non-local algorithms implemented on software requiring large computing time and resources. This point limits the use of RICH detectors in the trigger system of large experiments such as LHCb. An alternative solution, based on a cellular approach, is proposed. A list of possible templates from the standard library is established and considerations on the requirements for the hardware implementation in FPGAs are given. From the conceptual point of view, the cellular technique solution looks promising while the hardware implementation in FPGAs lays in the verge of the actual technical limitations if response times are to be in the order of the microsecond as required in the LHCb hardware trigger.
Two neuron CNN for hypothesis testing Vinyoles-Serra, M.; Vilasis-Cardona, X.
2012 13th International Workshop on Cellular Nanoscale Networks and their Applications,
2012-Aug.
Conference Proceeding
The two neuron continues time cellular neural network is used to define a statistic in the classical hypothesis testing problem. The proposal is based on a generalisation of the linear Fisher ...discriminant. The procedure to set the cellular neural network parameters is described and the performance shown on two examples with gaussianly distributed hypothesis. This technique might also be applied to probabilistic classification problems or pattern recognition.
Visual learning with cellular neural networks Badalov, A.; Vilasis-Cardona, X.; Albo-Canals, J.
2012 13th International Workshop on Cellular Nanoscale Networks and their Applications,
2012-Aug.
Conference Proceeding
Reinforcement learning is a powerful tool for teaching robotic agents to perform tasks in real environments. Visual information provided by a camera could be a cheap and rich source of information ...about an agent's surroundings, if this information were represented in a compact and generalizable form. We turn to cellular neural networks as the means of transforming visual input to a representation suitable for reinforcement learning. We investigate a CNN-based image processing algorithm and describe a method for efficiently computing CNNs using the DirectX 10 API.
Cellular Neural Networks for high energy physics Vilasis-Cardona, X.
2010 12th International Workshop on Cellular Nanoscale Networks and their Applications (CNNA 2010),
2010-Feb.
Conference Proceeding
Cellular Neural Networks (CNN) 1 main assets are quoted to be their capacity for parallel hardware implementation and their universality. On top, the possibility to add the information of a local ...sensor on every cell, provides a unique system for massive parallel signal processing responding in hardware time. Image processing has been, for a long time, the main field where the community has focussed its efforts to prove the excellence of CNNs. And, still, they are not used at large scale for image applications, probably because few cases are so demanding in terms of computation complexity and short response time not to be afforded by a standard sequential CPU
In this paper, we present the functionality of a low-cost camera sensor based on FPGAs. It is intended to be a solution for affordable robotic platforms requiring a smart vision sensor. This implies ...using simple image processing algorithms that are, nevertheless, flexible enough to ensure the robot navigation. The image processing elements and the connectivity both to the camera and the robot are embedded in an Actel Igloo FPGA, which fits the low power consumption, reprogrammability and cost requirements. We also present two applications. The first one, related to path planning, shows how a robot with such a device is able to identify different scenarios in the process of learning how to escape from a maze using reinforcement learning. The second application is a set of interactive activities using a companion robot and requiring robot vision devised to improve the recovery of children with brain trauma. In this particular case, both the cost and power consumption requirements for the camera sensor are demanding, since the robot has to be distributed to a large number of children and low consumption is essential to keep the robot working time within the therapeutic needs.
This paper gives necessary and sufficient conditions for the existence of limit cycles for an autonomous antisymmetric cellular neural network. The method combines the key arguments of the ...Poincare-Bendixon theorem with index theory. Further, the generalisation to the non-symmetric case is discussed.
The two neuron CNN system with piecewise linear activation function and time dependent external inputs is analyzed and the parametric solutions for the trajectories are given. We look then at the ...particular case in which the CNN has an antisymmetric template and sine external input. For particular parameter values, this example has been found in the literature to have chaotic behavior. We use the trajectories solution to study the sensitivity of the system to initial conditions and the possibility of self-intersecting trajectories. Both results characterize a chance for a chaotic behavior.
In this paper we give a proof of the universality of the Cellular Neural Network - Universal Machine (CNN-UM) alternative to those presented so far. On the one hand, this allows to find a general ...structure for CNN-UM programs; on the other hand, it helps to formally demonstrate that machine learning techniques can be used to find CNN-UM programs automatically. Finally, we report on two experiments in which our system is able to propose new efficient solutions.