A Multi-Agent Emotion engine
Emotions have always been known to play a very important and complex part in human behavior and no study of human nature is complete without mentioning them. Using a multi-agent emotion engine we try to generate artificial emotions in a way akin to their biological counterparts and use them for control applications. Rather than call them as synthetic/artificial emotions we prefer to designate them as an emotional control juice. Here each individual agent represents and generates a given emotion. Positive emotion generating agents suppress their counterparts while stimulate their own kind. Negative emotion generating agents do likewise. The tussle between these agents results in an internal mood within which serves to produce an adrenalin-like effect. The artificial adrenalin makes the engine sample the external world proportional to its production thereby attempting to bring the emoting system to a stable state. Life for stimulations and suppressions, a look-back factor, emotion decay and more importantly an emotional resource render the much needed natural and dynamic effects to the synthetic emotions churned by the engine.
A complete description of the multi-agent emotion engine that was simulated using LPA Prolog’s Chimera Agent System can be found HERE.
Learning Behaviours Through Facial Expression
We are also experimenting on how facial expressions can aid even the physically challenged to train or make a robot learn to perform a desired task. This research is a by-product of facial emotion recognition.The human face is detected and analyzed for its facial expression/emotional intensity which in turn serves as a reward/penalty to a robot thereby reinforcing/discouraging it from performing a sequence of tasks. The learning algorithm gradually sequences the actions performed by the robot thereby making it complete the desired task as also remember them.