During the movement of the mobile robot, except the common scenario, sometimes, the robot has to face the unexpected uncertainty in the environment. Traditional methods to address this problem generally use the task-specific method from the engineering perspective, lacking flexibility in the changing environment and being difficult to respond to the environmental challenges. Because of the function of the brain’s neuromodulatory system, human beings have ability to respond to the ever-changing environment quickly. To simulate the working mechanism of the human brain in responding to the unexpected uncertainty in the environment, this article presents a motivated developmental network (MDN) to offer a control configuration for an artificial agent to face the unexpected uncertainty in the environment, through introducing the acetylcholine/norepinephrine (ACh/NE) systems to the MDN. Taking the regulation role of ACh/NE in serotonin and dopamine (DA) into account, a novel learning rate for the hidden layer neurons of the MDN is proposed. Moreover, a novel composite mechanism is presented to decide the moving direction of the agent. Under the modulation of DA, serotonin, ACh, and NE, the agent can perform specific functions effectively, e.g., chase a target and elude the obstacle, especially the sudden obstacle. Goal-directed pursuing behavior in three simulation cases illustrates the effect of the presented neural modulatory systems, for instance, dealing with the unexpected uncertainty, realizing the attentional effort, reinforcement learning, etc. To the best of our knowledge, this article is the first endeavor to address the unexpected uncertainty with the MDN.
Read full abstract