Fig 2 - uploaded by Claes Strannegård
Content may be subject to copyright.
Left: A three-species predator-prey ecosystem, with grass, deer and wolves. The red bars show happiness values. Right: Population dynamics from two different simulations.
Source publication
We start by discussing the link between ecosystem simulators and artificial general intelligence (AGI). Then we present the open-source ecosystem simulator Ecotwin, which is based on the game engine Unity and operates on ecosystems containing inanimate objects like mountains and lakes, as well as organisms, such as animals and plants. Animal cognit...
Contexts in source publication
Context 1
... study concerns a three species predator-prey ecosystem with grass, deer, and wolves, as illustrated in Fig. 2 (Left). Deer and wolves have vision, which is modeled via Unity's ray casts, and gives the direction and distance to the closest visible objects of each type, within a certain radius [10]). Moreover, the deer can smell grass and wolves, while the wolves can smell deer. At each time step, each animal decides whether it should stand still, ...
Context 2
... ran two simulations with Ecotwin, starting with the same ecosystem. The result is shown in Fig. 2 (Right). As expected, given the dependence on randomness, the two simulations are different. In both simulations we see LotkaVolterra cycles, with an increase in grass, followed by an increase in deer, followed by an increase in wolves, and similar decreases. More details about the study can be found in [19]. ...
Citations
We present an AI-based ecosystem simulator that uses three-dimensional models of the terrain and animal models controlled by deep reinforcement learning. The simulations take place in a game engine environment, which enables continuous visual observation of the ecosystem model. The terrain models are generated from geographic data with altitudes and land cover type. The animal models combine three-dimensional conformation models with animation schemes and decision-making mechanisms trained with deep reinforcement learning in increasingly complex environments (curriculum learning). We show how AI tools of this kind can be used for modeling the development of specific ecosystems with and without different forms of economic activities. In particular, we show how they might be used for modeling local biodiversity effects of land cover change, exploitation of natural resources, pollution, invasive species, and climate change.
