Emotional Intervention on an Action Selection Mechanism Based on Artificial Immune Networks for Navigation of Autonomous Agents

This article investigates the effects of emotional intervention on artificial immune networks used for navigation of autonomous agents (simulated autonomous mobile robots). It is known from psychoneuroimmunology that stress influences the immune system response. From the various models of emotions related to stress available in literature, the computational model of the amygdala reported by Mochida, Ishiguro, Aoki, and Uchikawa (1995) is used in this article. The emotional intervention is implemented as a frustration signal coming from an artificial amygdala that influences the dynamics of antibody selection. A series of experiments with an autonomous agent implementing a collision-free goal-following behavior is presented in five simulated environments with different levels of difficulty. Two types of immune network based action selection mechanism are examined: (a) independently acting and (b) emotionally influenced. They are compared with each other in MATLAB simulations; their performance is estimated on the basis of time steps and their success in collision-free goal attainment. The artificial emotion mechanism modifies the immune response to overcome some difficult situations and to improve the performance of the behavior arbitration as a whole.