A Comparison of Genetic Algorithms and Reinforcement Learning for Optimising Sustainable Forest Management

Sustainable forest management is defined as “the stewardship and use of forests and forest lands in a way, and at a rate, that maintains their biodiversity, productivity, regeneration capacity, vitality and their potential to fulfil, now and in the future, relevant ecological, economic and social functions [...].” (MCPFE, 1994). As such, forest management has to satisfy multiple and often conflicting goals. Furthermore, forest planning is characterised by the long-term horizon of its outcomes. Since long-term plans are made in the face of uncertain futures, long-term sustainable forest management should incorporate some measure of risk. Uncertainty emerges from a variety of sources, including irregular or unknown fluctuations in the demand for timber, or the occurrence of extreme events. In addition, forest management is dynamic in time and space, for example, different stands have different properties, and the likelihood of stochastic events may change over time. Forest planning may be suboptimal if it ignores these sources of uncertainty and risk. Previous work on multi-objective optimisation in forest management has mainly used heuristic search methods. For example, Bettinger et al. (2002), Pukkala and Kurttila (2005) compare various heuristic optimisation techniques and conclude that Genetic Algorithms (GAs) perform well for more complex spatial problems. However, the studies did not investigate the algorithms' performance under uncertainty. Reinforcement Learning (RL) is an alternative approach for optimal policy selection. RL is a Machine Learning approach frequently used with agent-based systems (Sutton and Barto, 1998). Contemporary research using RL in the context of forest management has shown that it can find robust optimal solutions to multi-objective forest management problems, e.g. (Bone and Dragicevic, 2009). To further explore the potentials that RL provides over heuristic optimisation approaches, we perform a systematic comparison between RL and GA for sustainable forest management for tasks with increasing uncertainty.