Skip to Main Content
A learning model is presented to evolve the collective behavior of a group of autonomous entities in virtual reality applications. The entities are autonomous in that they have no means of direct communication with each other. The learning model uses a module called a “cognitive map”. The state of advancement of a task being performed by the group is given by this map. The group receives a reward according to the evolution of the task performed. Through an egalitarian distribution of the reward within the group, an automatic evolution of cooperation is attempted in order to complete the task successfully. At each time interval, an entity must choose its behavior from a basis set of options. Rewards the entity has received both in the distant and in the recent past influence the choice made by the entity.