Authors: Kiam, Jane Jean; Sabbadin, Régis;Ponzoni Carvalho Chanel, Caroline ;Harder, Benedikt;Schulte, Axel;

Given the scarcity of available information on poachers’ activities in vast conservation areas, game theoretical methods can assist rangers in devising strategies to curb illegal poaching, while considering the poachers’ decision-making behaviours. We introduce a formal stochastic game model, along with a solution approach interleaving dynamic programming and Nash equilibrium computation. Rangers and poachers are assumed to be non-cooperative players who make decisions simultaneously. Different from previous works (some use for example Stackelberg game), our stochastic game model considers the probabilistic dynamics of the game situation, modelled as a Markov decision process. Furthermore, we analyse results obtained from benchmarking against an iterative strategic game approach, in which normal-form games are solved at each decision-making round, without considering the probabilistic dynamics. We also analyse the performance of our solution method on games of varying size, rewards definition, game topology (considering sites with different geographical and ecological features) and players’ mixed capabilities (e.g. aerial support for rangers). Finally, we outline possible extensions of our work, aimed at lifting some assumptions by considering i) imperfect state observation, ii) more complex interactions between poachers, rangers and animals' dynamics, as well as iii) players' different subjective preferences.