Artificial Gene Regulatory Networks for Agent Control: Iba/Evolutionary

Sylvain Cussat-Blanc (1, 2) , Jean Disset (1) , Stephane Sanchez (1, 2) , Yves Duthen (1, 2)

1 IRIT-REVA - Real Expression Artificial Life
2 UT Capitole - Université Toulouse Capitole

This chapter shows how an artificial gene regulatory network (GRN) can be used to control various kinds of virtual agents. It first details the computational model of the GRN and shows how one can evolve it with a genetic algorithm. Next, it reviews the behavioral abilities of the GRN with images, which show the variation of output expressions according to input values, and with videos, which add the temporal aspect of gene regulation. The chapter then presents the use of this GRN to control three kinds of agents. The first experiment shows its application to control virtual cells in an artificial embryogenesis process. Then, the same GRN is directly connected to a virtual race car and is taught to drive on a track. Finally, the last experiment uses the GRN to regulate a set of high‐level scripted behaviors for a team of agents to defend a target against incoming threats.

Domaines: Intelligence artificielle [cs.AI]

Sylvain Cussat-Blanc, Jean Disset, Stephane Sanchez, Yves Duthen. Artificial Gene Regulatory Networks for Agent Control: Iba/Evolutionary. Hitoshi, Iba; Nasimul, Noman. Evolutionary Computation in Gene Regulatory Network Research, 12, Wiley, pp.299-326, 2016, Wiley Series in Bioinformatics, 978-1-118-91151-8. ⟨10.1002/9781119079453.ch12⟩. ⟨hal-03159058⟩