Exploring a set of unlabeled data to extract the similar clusters, known as data clustering, is an appealing problem in machine learning. In other words, data clustering organizes the underlying data into different groups using a notion of similarity between patterns. A new approach to solve the data clustering problem based on multiple cooperative swarms is introduced. The proposed approach is inspired by the social swarming behavior of biological bird flocks which search for food situated in several places. The proposed approach is composed of two main phases, namely, initialization and exploitation. In the initialization phase, the aim is to distribute the search space among several swarms. That is, a part of the search space is assigned to each swarm in this phase. In the exploitation phase, each swarm searches for the center of its associated cluster while cooperating with other swarms. The search proceeds to converge to a near-optimal solution. As compared to the single swarm clustering approach, the proposed multiple cooperative swarms provide better solutions in terms of fitness function measure for the cluster centers, as the dimensionality of data and number of clusters increase. The multiple cooperative swarms clustering approach assumes that the number of clusters is known a priori. The notion of stability analysis is proposed to extract the number of clusters for the underlying data using multiple cooperative swarms. The mathematical explanations demonstrating why the proposed approach leads to more stable and robust results than those of the single swarm clustering are also provided. Application of the proposed multiple cooperative swarms clustering is considered for one of the most challenging problems in speech recognition: phoneme recognition. The proposed approach is used to decompose the recognition task into a number of subtasks or modules. Each module involves a set of similar phonemes known as a phoneme family. Basically, the goal is to obtain the best solution for phoneme families using the proposed multiple cooperative swarms clustering. The experiments using the standard TIMIT corpus indicate that using the proposed clustering approach boosts the accuracy of the modular approach for phoneme recognition considerably.

Exploring a set of unlabeled data to extract the similar clusters, known as data clustering, is an appealing problem in machine learning. In other words, data clustering organizes the underlying data into different groups using a notion of similarity between patterns. A new approach to solve the data clustering problem based on multiple cooperative swarms is introduced. The proposed approach is inspired by the social swarming behavior of biological bird flocks which search for food situated in several places. The proposed approach is composed of two main phases, namely, initialization and exploitation. In the initialization phase, the aim is to distribute the search space among several swarms. That is, a part of the search space is assigned to each swarm in this phase. In the exploitation phase, each swarm searches for the center of its associated cluster while cooperating with other swarms. The search proceeds to converge to a near-optimal solution. As compared to the single swarm clustering approach, the proposed multiple cooperative swarms provide better solutions in terms of fitness function measure for the cluster centers, as the dimensionality of data and number of clusters increase. The multiple cooperative swarms clustering approach assumes that the number of clusters is known a priori. The notion of stability analysis is proposed to extract the number of clusters for the underlying data using multiple cooperative swarms. The mathematical explanations demonstrating why the proposed approach leads to more stable and robust results than those of the single swarm clustering are also provided. Application of the proposed multiple cooperative swarms clustering is considered for one of the most challenging problems in speech recognition: phoneme recognition. The proposed approach is used to decompose the recognition task into a number of subtasks or modules. Each module involves a set of similar phonemes known as a phoneme family. Basically, the goal is to obtain the best solution for phoneme families using the proposed multiple cooperative swarms clustering. The experiments using the standard TIMIT corpus indicate that using the proposed clustering approach boosts the accuracy of the modular approach for phoneme recognition considerably.

Optimization is a mathematical technique that concerns the finding of maxima or minima of functions in some feasible region. There is no business or industry which is not involved in solving optimization problems. A variety of optimization techniques compete for the best solution. Particle Swarm Optimization (PSO) is a relatively new, modern, and powerful method of optimization that has been empirically shown to perform well on many of these optimization problems. It is widely used to find the global optimum solution in a complex search space. This thesis aims at providing a review and discussion of the most established results on PSO algorithm as well as exposing the most active research topics that can give initiative for future work and help the practitioner improve better result with little effort. This paper introduces a theoretical idea and detailed explanation of the PSO algorithm, the advantages and disadvantages, the effects and judicious selection of the various parameters. Moreover, this thesis discusses a study of boundary conditions with the invisible wall technique, controlling the convergence behaviors of PSO, discrete-valued problems, multi-objective PSO, and applications of PSO. Finally, this paper presents some kinds of improved versions as well as recent progress in the development of the PSO, and the future research issues are also given.

Swarm Intelligence (SI) is a young field of study from which solutions to complex problems have been proposed based on how some natural organisms (e.g. ants, bees and others) achieve many of their daily tasks through simple sets of interactions. This thesis proposes two models for area exploration and coverage based on SI principles. These two models present a novel approach based on the combination of: ants’ pheromones, in order to keep track of visited places; and bird flocks or fish schooling, so as to move and collaborate. An implementation of both models was done in order to simulate and evaluate both the emergent behavior of the agents as well as their area exploration and coverage performance. Based on the outcome of the simulations it is concluded that both models are able to perform the exploration and coverage task and that one model is better than the other.

Scheduling of aperiodic and independent tasks in hard real-time symmetric multiprocessing systems is an NP-complete problem that is often solved using heuristics like particle swarm optimization (PSO). The performance of these class of heuristics, known as evolutionary algorithms, are often evaluated based on the number of iterations it takes to find a solution. Such metrics provide limited information on how the algorithm reaches a solution and how the process could be accelerated. This thesis presents a methodology to analyze the trajectory formed by candidate solutions in order to analyze them in both the time and frequency domains at a single scale. The analysis entails (i) the impact of different parameters for the PSO algorithm, and (ii) the evolutionary processes in the swarm. The work reveals that particles have a directed movement towards a solution during a transient phase, and then enter a steady state where they perform an unguided local search. The scheduling algorithm presented in this thesis uses a variation of the minimum total tardiness with cumulative penalties cost function, that can be extended to suit different system needs. The experimental results show that the scheduler is able to distribute tasks to meet the real-time deadlines over 1, 2, and 4 processors and up to 30 tasks with overall system loads of up to 50\% in fewer than 1,000 iterations. When scheduling greater loads, the scheduler reaches local solutions with 1 to 2 missed deadlines, while larger tasks sets take longer to converge. The trajectories of the particles during the scheduling algorithm are examined as a means to emphasize the impact of the behaviour on the application performance and give insight into ways to improve the algorithm for both space and terrestrial applications.

Scheduling of aperiodic and independent tasks in hard real-time symmetric multiprocessing systems is an NP-complete problem that is often solved using heuristics like particle swarm optimization (PSO). The performance of these class of heuristics, known as evolutionary algorithms, are often evaluated based on the number of iterations it takes to find a solution. Such metrics provide limited information on how the algorithm reaches a solution and how the process could be accelerated. This thesis presents a methodology to analyze the trajectory formed by candidate solutions in order to analyze them in both the time and frequency domains at a single scale. The analysis entails (i) the impact of different parameters for the PSO algorithm, and (ii) the evolutionary processes in the swarm. The work reveals that particles have a directed movement towards a solution during a transient phase, and then enter a steady state where they perform an unguided local search. The scheduling algorithm presented in this thesis uses a variation of the minimum total tardiness with cumulative penalties cost function, that can be extended to suit different system needs. The experimental results show that the scheduler is able to distribute tasks to meet the real-time deadlines over 1, 2, and 4 processors and up to 30 tasks with overall system loads of up to 50\% in fewer than 1,000 iterations. When scheduling greater loads, the scheduler reaches local solutions with 1 to 2 missed deadlines, while larger tasks sets take longer to converge. The trajectories of the particles during the scheduling algorithm are examined as a means to emphasize the impact of the behaviour on the application performance and give insight into ways to improve the algorithm for both space and terrestrial applications.

Population-based global optimization algorithms including Particle Swarm Optimization (PSO) have become popular for solving multi-optima problems much more efficiently than the traditional mathematical techniques. In this research, we present and evaluate a new parallel PSO algorithm that provides a significant performance improvement as compared to the serial PSO algorithm. Instead of merely assigning parts of the task of serial version to several processors, the new algorithm places multiple swarms on the available nodes in which operate independently, while collaborating on the same task. With the reduction of the communication bottleneck as well the ability to manipulate the individual swarms independently, the proposed approach outperforms the original PSO algorithm and still maintains the simplicity and ease of implementation.

The Ant Colony Optimization and Particle Swarm Optimization are two Swarm Intelligence algorithms often utilized for optimization. Swarm Intelligence relies on agents that possess fragmented knowledge, a concept not often utilized in games. The aim of this study is to research whether there are any benefits to using these Swarm Intelligence algorithms in comparison to standard algorithms such as A* for pathfinding in a game. Games often consist of dynamic environments with mobile agents, as such all experiments were conducted with dynamic destinations. Algorithms were measured on the length of their path and the time taken to calculate that path. The algorithms were implemented with minor modifications to allow them to better function in a grid based environment. The Ant Colony Optimization was modified in regards to how pheromone was distributed in the dynamic environment to better allow the algorithm to path towards a mobile target. Whereas the Particle Swarm Optimization was given set start positions and velocity in order to increase initial search space and modifications to increase particle diversity. The results obtained from the experimentation showcased that the Swarm Intelligence algorithms were capable of performing to great results in terms of calculation speed, they were however not able to obtain the same path optimality as A*. The algorithms' implementation can be improved but show potential to be useful in games.

Techniques for algorithmic musical composition or generative music working directly with the frequencies of the sounds being played are rare today as most approaches rely on mapping of discrete states. The purpose of this work is to investigate how self organizing audio can be created in realtime based on pitch information, and to find methods that give both expressive control and some unpredictability. A series of experiments were done using SuperCollider and evaluated against criteria formulated using music theory and psychoacoustics. One approach was utilizing the missing fundamental phenomenon and pitch detection using autocorrelation. This approach generated unpredictable sounds but was too much reliant on user input to generate evolving sounds. Another approach was the Kuramoto model of synchronizing oscillators. This resulted in pleasant phasing sounds when oscillators modulating the amplitudes of audible oscillators were synchronized, and distorted sounds when the frequencies of the audible oscillators were synchronized. Lastly, swarming behavior was investigated by implementing an audio analogy of Reynolds’ Boids model. The boids model resulted in interesting independently evolving sounds. Only the boids model showed true promise as a method of algorithmic composition. Further work could be done to expand the boids model by incorporating more parameters. Kuramoto synchronization could viably be used for sound design or incorporated into the boids model.

By enabling coordinated task execution and movement, robotic swarms can achieve efficient exploration or disaster site management. When utilizing Ultra-wideband (UWB) radio technology for ranging, the proposed relative localization sensor can be made lightweight and relatively indifferent to the ambient environment. Infrastructure dependency is eliminated by making the whole sensor fit on a swarm agent, while allowing for a certain amount of positional error. In this thesis, a novel algorithm is implemented in to constrained hardware and compared to a more traditional trilateration approach. Both algorithms utilize Particle Swarm Optimization (PSO) to be more robust towards noise and achieves similar accuracy, but the proposed algorithm can run up to ten times faster. The antenna array which forms the localization sensor weighs only 56g, and achieves around 0.5m RMSE with a 10Hz update rate. Experiments show that the accuracy can be further improved if the rotational bias observed in the UWB devices are compensated for.