Search

The research field of Artificial General Intelligence (AGI) is concerned with the creation of adaptive programs that can autonomously address tasks of a different nature. Search and planning have been identified as core capabilities of AGI, and have been successful in many scenarios that require sequential decision-making. However, many search algorithms are developed for specific problems and exploit domain-specific knowledge, which makes them not applicable to perform different tasks autonomously. Although some domain-independent search algorithms have been proposed, a programmer still has to make decisions on their design, setup and enhancements. Thus, the performance is limited by the programmer's decisions, which are usually biased. This paper proposes to develop a framework that, in line with the goals of AGI, autonomously addresses a wide variety of search tasks, adapting automatically to each new, unknown task. To achieve this, we propose to encode search algorithms in a formal language and combine algorithm portfolios with automatic algorithm generation. In addition, we see games as the ideal test bed for the framework, because they can model a wide variety of complex problems. Finally, we believe that this research will have an impact not only on the AGI research field, but also on the game industry and on real-world problems.

Monte Carlo Tree Search (MCTS) is a heuristic search algorithm that can play a wide range of games without requiring any domain-specific knowledge. However, MCTS tends to struggle in very complicated games due to an exponentially increasing branching factor. A promising solution for this problem is to focus the search only on a small fraction of states. Subgoal Monte Carlo Tree Search (S-MCTS) achieves this by using a predefined subgoal-predicate that detects promising states called subgoals. However, not only does this make S-MCTS domaindependent, but also it is often difficult to define a good predicate. In this paper, we propose using quality diversity (QD) algorithms to detect subgoals in real-time. Furthermore, we show how integrating QD-algorithms into S-MCTS significantly improves its performance in the Physical Travelling Salesmen Problem without requiring any domain-specific knowledge.

For general video game playing agents, the biggest challenge is adapting to the wide variety of situations they encounter and responding appropriately. Some success was recently achieved by modifying search-control parameters in agents on-line, during one play-through of a game. We propose adapting such methods for Rolling Horizon Evolutionary Algorithms, which have shown high performance in many different environments, and test the effect of on-line adaptation on the agent's win rate. On-line tuned agents are able to achieve results comparable to the state of the art, including first win rates in hard problems, while employing a more general and highly adaptive approach. We additionally include further insight into the algorithm itself, given by statistics gathered during the tuning process and highlight key parameter choices.

A purpose of General Video Game Playing (GVGP) is to create agents capable of playing many different real-time video games. Instead of using a fixed general strategy, a challenging aspect is devising strategies that adapt the search to each video game being played. Recent work showed that on-line parameter tuning can be used to adapt Monte-Carlo Tree Search (MCTS) in real-time. This paper extends prior work on Self-adaptive Monte-Carlo Tree Search (SA-MCTS) by further testing one of the previously proposed on-line parameter tuning strategies, based on the N-Tuple Bandit Evolutionary Algorithm (NTBEA). Results show that, both for a simple and a more advanced MCTS agent, on-line parameter tuning has impact on performance only for a few GVGP games. Moreover, an informed strategy as NTBEA shows a significant performance increase only in one case. In a real-time domain as GVGP, advanced parameter tuning does not seem very promising. Randomizing pre-selected parameters for each simulation appears to be a robust approach.

General Video Game Playing (GVGP) is a field of Artificial Intelligence where agents play a variety of real-time video games that are unknown in advance. This limits the use of domain-specific heuristics. Monte-Carlo Tree Search (MCTS) is a search technique for game playing that does not rely on domain-specific knowledge. This paper discusses eight enhancements for MCTS in GVGP; Progressive History, N-Gram Selection Technique, Tree Reuse, Breadth-First Tree Initialization, Loss Avoidance, Novelty-Based Pruning, Knowledge-Based Evaluations, and Deterministic Game Detection. Some of these are known from existing literature, and are either extended or introduced in the context of GVGP, and some are novel enhancements for MCTS. Most enhancements are shown to provide statistically significant increases in win percentages when applied individually. When combined, they increase the average win percentage over sixty different games from 31.0% to 48.4% in comparison to a vanilla MCTS implementation, approaching a level that is competitive with the best agents of the GVG-AI competition in 2015.