Yeni hücresel doğrusal olmayan ve hücresel lojik ağların gerçeklemeleri ve uygulamaları

This thesis is a consistent and coherent reorganization of studies on two topics of nonlinear systems. First topic includes Relaxation Oscillators and logic oscillators with similar behavior which are locally coupled and the resulting Cellular Nonlinear Networks (CNN) are utilized for a predictive motion planning algorithm. Nonlinear waves, especially autowave and traveling wave, have been studied and their system model, coupling schemes, parameters, and inputs generating both types of nonlinear waves are explained. The research covers two implementations of selected CNN and compares their digital circuit (FPGA prototyping), CPU simulation and GPU simulation performances. The research is focused on the Doppler Effect occurrence of the propagated nonlinear waves. A novel nonlinear wave propagation based feedback motion planning algorithm which utilizes the Doppler Effect and generates a prediction for the future state of target object has been proposed. The comparisons which reveals the effect of Doppler Effect are reported. The results prove that a tracker even slower than the target may catch it using the proposed algorithm. This new method of motion planning needs two layers of oscillator based CNNs. Two types of relaxation oscillators (one of them is a new model) and the logic oscillator have been tested for the algorithm. Novel models of chaotic time-delay systems are introduced in the thesis as the second topic. The proposed binary output nonlinearity makes the oscillator generate a mono-scroll chaotic attractor. This thesis also proposes a generalization of the binary output nonlinear function, which is a quantized output nonlinearity. The generalized nonlinearity yields a multi-scroll attractor. Both systems are modelled as sampled-data models, because the binary delay lines are constructed by digital components (D-type flip-flops). The research on implementations of these oscillators has been expanded with binary inverting buffers (NOT gates) and asynchronous digital state machines. These systems successfully generate true random bit sequences without the need for post-processing. Up-to-date NIST's statistical test suite is used for the tests of bit sequences and successful throughput rates are reported. The jitter on the NOT gate based delay line is utilized as physical noise and all-digital implementation supported by the jitter also passed the statistical tests. The thesis merges research parts and reorganize the outputs under four titles: cells, networks, implementations and applications.