Cloud computing for synergised emotional model evolution in multi-agent learning systems

Machine learning is a technology paramount to enhancing the adaptability of agent-based systems. Learning is a desirable aspect in synthetic characters, or 'believable ' agents, as it offers a degree of realism to their interactions. The advantage of collaborative efforts in multi-agent learning systems can be overshadowed by concerns over system scalability and adaptive dynamics, particularly when learning capabilities further add dynamics and introduce mathematical anomalies. The proposed Multi-agent Learning through Distributed Artificial Consciousness (MALDAC) Architecture is a scalable approach to developing adaptable systems in complex, believable environments. The MALDAC Architecture uses cloud computing and multi-agent learning. It applies emotional models and artificial consciousness theory to mitigate scalability issues whilst coping with system dynamics. The architecture consists of Context-based Adaptive Emotions Driven Agents (CAEDA) which apply adaptive consciousness and emotional model processing to collaboratively develop and adapt agents' behaviour. CAEDA agents selectively aggregate consciousness-based functional modules provided via web service agents. A virtual environment implementing the MALDAC architecture is shown to enhance scalability in multi-agent learning systems, particularly in stochastic and dynamic environments.