In recent years researchers have shown increasing interest in swarm intelligence as a promising approach to adaptive distributed problem solving. Swarm intelligence consists of techniques inspired by nature, especially social insects and aggregations of animals, and even human interactions. They are based on self-organization (a system's overall behavior emerges from the local interactions among its relatively simple components) and are often decentralized and massively distributed. Particle systems are an approach to swarm intelligence that focus on collective movements, and have been used successfully for applications such as computer animation in graphics and control of movements of autonomous robotic vehicle teams. However, particle system techniques have not been applied substantially to problem solving beyond merely collective navigational tasks.

In this dissertation, I present an extension to particle systems that incorporates top-down, high-level control to self-organizing mobile agents, thereby guiding the self-organizing process and making it possible for particle systems to undertake problem solving directed by goal-oriented behavior while retaining their decentralized, local nature. This extended particle system approach is critically evaluated through three experimental studies that are adapted from well-known problems in multi-agent systems: search and collect, cooperative transport and logistics. The results provide evidence that extended particle systems are capable of exhibiting behavior important for distributed problem solving, such as cooperative sensing, division of labor, sharing of information, and developing global strategies through local interactions. They also show that aggregated movements can be utilized to create coordination at different levels and phases of the performance of a task, whether those include navigation or not, making extended particle systems a useful tool in the construction of adaptive distributed systems.