EAs + Monte Carlo Tree Search for the win (*BEST PAPER of EvoAPPS 2020 conference*)

Yes! We did it! :D

Yesterday we presented our last paper on the scope of AI in Videogames entitled «Testing hybrid computational intelligence algorithms for general game playing«, by two colleagues of the University of Málaga (A. E. Vázquez-Núñez and A. J. Fernández-Leiva), together with P. García-Sánchez (@fergunet) and I (A. M. Mora).

It was focused on the domain of General Game Playing, where the aim is to create an autonomous agent able to adapt itself to play to several different videogames (not previously known for it). We used the framework and followed the rules of the General Video Game AI Competition.

Thus, it’s an approach to implement a kind of human-like behaviour the first time we face a new game.

The manuscript was selected as the Best Paper of EvoAPPS 2020 conference (inside EVO* 2020), held online for the first time, due to COVID-19 situation.

The abstract of the work is:

General Videogame Playing is one of the hottest topics in the research field of AI in videogames. It aims at the implementation of algorithms or autonomous agents able to play a set of unknown games efficiently, just receiving the set of rules to play in real time. Thus, this work presents the implementation of eight approaches based on the main techniques applied in the literature to face this problem, including two different hybrid implementations combining Monte Carlo Tree Search and Genetic Algorithms. They have been created within the General Video Game Artificial Intelligence (GVGAI) Competition platform. Then, the algorithms have been tested in a set of 20 games from that competition, analyzing its performance. According to the obtained results, we can conclude that the proposed hybrid approaches are the best approaches, and they would be a very competitive entry for the competition.

And here is the presentation:

Enjoy it!

And, as usual, cite us. ;D

A new lap in the race to improve our evolutionary fuzzy-controllers for TORCS

Last week we presented at the IEEE Conference on Game 2019, held in London (UK), our new paper titled «Beating uncertainty in racing bot evolution through enhanced exploration and pole position selection«.

The abstract of the work is:

One of the main problems in the design through optimization of car racing bots is the inherent noise in the optimization process: besides the fact that the fitness is a heuristic
based on what we think are the keys to success and as such just a surrogate for the ultimate objective, winning races, fitness itself is uncertain due to the stochastic behavior of racing conditions and the rest of the (simulated) racers. The fuzzy-based genetic controller for the car racing simulator TORCS that we have defined in previous works is based on two fuzzy subcontrollers, one for deciding on the wheel steering angle and another to set the car target speed at the next simulation tick. They are both optimized by means of an Evolutionary Algorithm, which considers an already tested fitness function focused on the maximization of the average speed during the race and the minimization of the car damage. The noisy environment asks for keeping diversity high during evolution, that is why we have added a Blend Crossover (BLX-alpha) operator, which is, besides, able to exploit current results at the same time it explores. Additionally, we try to address uncertainty in selection by introducing a novel selection policy of parents based in races, where the individuals are grouped and compete against others in several races, so just the firsts ranked will remain in the population as parents. Several experiments have been conducted, testing the value of the different controllers. The results show that the combination of a dynamic BLX-alpha crossover operator plus the pole position selection policy clearly beats the rest of approaches. Moreover, in the comparison of this controller with one of the participants of the prestigious international Simulated Car Racing Championship, our autonomous driver obtains much better results than the opponent.

The presentation can be seen below:

As usual, enjoy it and…cite us! :D

Making birds less angry by evolving pig supporting structures in the game

After our poster in EvoStar, GECCO 2019 saw another poster on evolution of Angry Birds structures, in this ocassion focused on the inclusion of the Box2D Physics simulation engine into the evolutionary algorithm to save using Science Birds, which improved evaluation of the structures that needed it by 100x.
The poster is minimalistic, with the intention of making it awesome.
//embedr.flickr.com/assets/client-code.js
Get data, code and the paper itself from our repository.

Improved Genetic Fuzzy Drivers presented at CIG 2018

Last week I presented at IEEE CIG 2018 (held in Maastricht, The Netherlands) our following step in our research about autonomous drivers for Car Racing Simulators, such as TORCS, titled «The Evolutionary Race: Improving the Process of Evaluating Car Controllers in Racing Simulators«.

As commented before by @jjmerelo and later by @fergunet, we designed with Mohammed Salem (University of Mascara) a driver’s AI in which two Fuzzy Subcontrollers were hybridized with a Genetic Algorithm.

In this work we present a better evaluation approach for the GA, combining three methods: heuristic track choosing, improved fitness functions, and race-based selection of the best.

The abstract of the work is:

Simulated car races have been used for a long time as an environment where car controlling algorithms can be tested; they are an interesting testbed for all kinds of algorithms, including metaheuristics such as evolutionary algorithms. However, the challenge in the evolutionary algorithms is to design a reliable and effective evaluation process for the individuals that eventually translates into good solutions to the car racing problem: finding a controller that is able to win in a wide range of tracks and with a good quantity of opponents. Evaluating individual car controllers involves not only the design of a proper fitness function representing how good the car controller would be in a competitive race, but also the selection of the best solution for the optimization problem being solved; this decision might not be easy when uncertainty is present in the problem environment; in this case, weather and track conditions as well as unpredictable behavior of other drivers. Creating a methodology for the automatic design of the controller of an autonomous driver for a car racing simulator such as TORCS is an optimization problem which offers all these challenges. Thus, in this paper we describe an analysis and some proposals to improve the evaluation of optimized fuzzy drivers for TORCS over previous attempts to do so. It builds on preliminary results obtained in previous papers as a baseline and aims to obtain a more competitive autonomous driver via redesign of the fitness evaluation procedure; to this end, two different fitness functions are studied in several experiments, along with a novel race-based approach for the selection of the best individual in the evolution.

And the presentation is:

You can check our paper in the proceedings of the conference.

Enjoy it!

(And cite us as usual :D)

On Volunteer-Computing and Self-driving car fuzzy controllers in the sunny Cádiz

Every two years, the International Conference on Information Processing and Management of Uncertainty in Knowledge-Based Systems (IPMU) brings together the most important researchers in the area of uncertainty and fuzzy systems. As I am working in Cadiz, it was a great opportunity to present some of the latest work that the Geneura group has recently developed.

The first of these has been developed together with the Technical Institute of Tijuana and describes the social behaviour of users of a voluntary computer system. It is very interesting to discover how the use of a leaderboard makes users spend more time collaborating. Take  a look to the presentation:

Mario García Valdez, Juan Julián Merelo Guervós, Lucero Lara, Pablo García-Sánchez:
Increasing Performance via Gamification in a Volunteer-Based Evolutionary Computation System. IPMU (3) 2018: 342-353

Here is the abstract:

Distributed computing systems can be created using volunteers, users who spontaneously, after receiving an invitation, decide to provide their own resources or storage to contribute to a common effort. They can, for instance, run a script embedded in a web page; thus, collaboration is straightforward, but also ephemeral, with resources depending on the amount of time the user decides to lend. This implies that the user has to be kept engaged so as to obtain as many computing cycles as possible. In this paper, we analyze a volunteer-based evolutionary computing system called NodIO with the objective of discovering design decisions that encourage volunteer participation, thus increasing the overall computing power. We present the results of an experiment in which a gamification technique is applied by adding a leader-board showing the top scores achieved by registered contributors. In NodIO, volunteers can participate without creating an account, so one of the questions we wanted to address was if the need to register would have a negative impact on user participation. The experiment results show that even if only a small percentage of users created an account, those participating in the competition provided around 90% of the work, thus effectively increasing the performance of the overall system.

 

The second work uses an evolutionary algorithm to optimize the parameters of a fuzzy controller that drives a car in the TORCS video game and continues our previous work. We have been collaborating with Mohammed Salem of University of Mascara along this line for a while.

Mohammed Salem, Antonio Miguel Mora, Juan Julián Merelo Guervós, Pablo García-Sánchez: Applying Genetic Algorithms for the Improvement of an Autonomous Fuzzy Driver for Simulated Car Racing. IPMU (3) 2018: 236-247

Games offer a suitable testbed where new methodologies and algorithms can be tested in a near-real life environment. For example, in a car driving game, using transfer learning or other techniques results can be generalized to autonomous driving environments. In this work, we use evolutionary algorithms to optimize a fuzzy autonomous driver for the open simulated car racing game TORCS. The Genetic Algorithm applied improves the fuzzy systems to set an optimal target speed as well as the instantaneous steering angle during the race. Thus, the approach offer an automatic way to define the membership functions, instead of a manual or hill-climbing descent method. However, the main issue with this kind of algorithms is to define a proper fitness function that best delivers the obtained result, which is eventually to win as many races as possible. In this paper we define two different evaluation functions, and prove that fine-tuning the controller via evolutionary algorithms robustly finds good results and that, in many cases, they are able to play very competitively against other published results, with a more relying approach that needs very few parameters to tune. The optimized fuzzy-controllers (one per fitness) yield a very good performance, mainly in tracks that have many turning points, which are, in turn, the most difficult for any autonomous agent. Experimental results show that the enhanced controllers are very competitive with respect to the embedded TORCS drivers, and much more efficient in driving than the original fuzzy-controller.

 

Creating Hearthstone decks by using Genetic Algorithms

I’m glad you’re here, friend! There’s a chill outside, so pull up a chair by the hearth of our inn and prepare to learn how the Ancient Gods use the power of the secret and ancient branch of the Evolution to generate Hearthstone decks by means of the magic and mistery!!

Several months ago, my colleague Alberto Tonda and I were discussing about our latest adventures playing the Digital Collectible Card Game Hearthstone, when one of us said «Uhm, Genetic Algorithms usually work well with combinatorial problems, and solutions are usually a vector of elements. Elements such as cards. Such as cards of Hearthstone, the game we are playing right now while we are talking. Are you thinking what I’m thinking?»

Five minutes later we found an open-source Hearthstone simulator and started to think how to address the possibility of automatically evolve decks of Hearthstone.

The idea is quite simple: Hearthstone is played using a deck of 30 cards (from a pool of thousands available), so it is easy to model the candidate solution. With the simulator, we can perform several matches using different enemy decks, and obtain the number of victories. Therefore, we have a number that can be used to model the performance (fitness) of the deck.

Soooo, it’s easy to see one and one makes two, two and one makes three, and it was destiny, that we created a genetic algorithm that generates deck for Hearthstone for free.

Our preliminary results where discussed here, but we wanted to continue testing our method, so we tested using all available classes of the game, with the help of JJ, Giovanny and Antonio. All the best human-made decks were outperformed by our approach! And not only that, we applied a new operator called Smart Mutation that it is based in what we do when we test new decks in Hearthstone: we remove a card, and place another instead, but with +/-1 mana crystals, and not one completely random from the pool. The results were even better. Neat!

Maybe you prefer to read the abstract, that it is written in a more formal way than this post. You know, using the language of the science.

Collectible card games have been among the most popular and profitable products of the entertainment industry since the early days of Magic: The Gathering in the nineties. Digital versions have also appeared, with HearthStone: Heroes of WarCraft being one of the most popular. In Hearthstone, every player can play as a hero, from a set of nine, and build his/her deck before the game from a big pool of available cards, including both neutral and hero-specific cards.
This kind of games offers several challenges for researchers in artificial intelligence since they involve hidden information, unpredictable behaviour, and a large and rugged search space. Besides, an important part of player engagement in such games is a periodical input of new cards in the system, which mainly opens the door to new strategies for the players. Playtesting is the method used to check the new card sets for possible design flaws, and it is usually performed manually or via exhaustive search; in the case of Hearthstone, such test plays must take into account the chosen hero, with its specific kind of cards.
In this paper, we present a novel idea to improve and accelerate the playtesting process, systematically exploring the space of possible decks using an Evolutionary Algorithm (EA). This EA creates HearthStone decks which are then played by an AI versus established human-designed decks. Since the space of possible combinations that are play-tested is huge, search through the space of possible decks has been shortened via a new heuristic mutation operator, which is based on the behaviour of human players modifying their decks.
Results show the viability of our method for exploring the space of possible decks and automating the play-testing phase of game design. The resulting decks, that have been examined for balancedness by an expert player, outperform human-made ones when played by the AI; the introduction of the new heuristic operator helps to improve the obtained solutions, and basing the study on the whole set of heroes shows its validity through the whole range of decks.

You can download the complete paper from the Knowledge-based Systems Journal https://www.sciencedirect.com/science/article/pii/S0950705118301953

See you in future adventures!!!

Early prediction of the outcome of Starcraft Games

As a result of Antonio Álvarez Caballero master’s thesis, we’ll be presenting tomorrow at the IJCCI 2017 conference a poster on the early prediction of Starcraft games.
The basic idea behind this line of research is to try and find a model of the game so that we can do fast fitness evaluation of strategies without playing the whole game, which can take up to 60 minutes. That way, we can optimize those strategies in an evolutionary algorithm and find the best ones.
In our usual open science style, paper and data are available in a repository.
Our conclusions say that we might be able to pull that off, using k-nearest neighbor algorithm. But we might have to investigate a bit further if we really want to find a model that gives us some insight about what makes a strategy a winner.

Dark clouds allow early prediction of heavy rain in Funchal, near where IJCCI is taking place

Our TORCS driving controller presented at EvoGAMES 2017

Last week, @jjmerelo presented at EvoGAMES 2017 (inside Evo* 2017) our work titled «Driving in TORCS using modular fuzzy controllers».

This paper presents a novel car racing controller for TORCS (The Open Racing Car Simulator), which is based in the combination of two fuzzy subcontrollers, one for setting the speed, and one to control the steer angle. The obtained results are quite promissing, as the controller is quite competitive even against very tough TORCS teams.

The abstract of the paper is:

«When driving a car it is essential to take into account all possible factors; even more so when, like in the TORCS simulated race game, the objective is not only to avoid collisions, but also to win the race within a limited budget. In this paper, we present the design of an autonomous driver for racing car in a simulated race. Unlike previous controllers, that only used fuzzy logic approaches for either acceleration or steering, the proposed driver uses simultaneously two fuzzy controllers for steering and computing the target speed of the car at every moment of the race. They use the track border sensors as inputs and besides, for enhanced safety, it has also taken into account the relative position of the other competitors. The proposed fuzzy driver is evaluated in practise and timed races giving good results across a wide variety of racing tracks, mainly those that have many turning points.»

There was an interactive presentation at the conference, together with a poster:

The paper is available online from: https://link.springer.com/chapter/10.1007/978-3-319-55849-3_24

Enjoy (and cite) it! :D

 

Master of Evolution! Using Genetic Algorithms to generate decks for the game HearthStone

This September we attended to the IEEE CIG 2017 Conference in Santorini, Greece, to present the paper «Evolutionary Deckbuilding in HearthStone». This paper was written in collaboration with our colleagues Alberto Tonda and Giovanni Squillero.

The story of this paper started a (not so) long time ago while me and Alberto were discussing about how awesome HearthStone is. Suddenly, we thought about how easy it would be to create the constraints for the uGP framework, and that there were some open source simulators of the game. In a while, we already had the constraints, the simulator adapted to accept individuals from uGP, and some experiments running.

And we finished the paper after, of course.

You can download the paper draft from here (the electronic original version is not available yet).

And here is the presentation:

The abstract:

—One of the most notable features of collectible card games is deckbuilding, that is, defining a personalized deck before the real game. Deckbuilding is a challenge that involves a big and rugged search space, with different and unpredictable behaviour
after simple card changes and even hidden information. In this paper, we explore the possibility of automated deckbuilding: a genetic algorithm is applied to the task, with the evaluation delegated to a game simulator that tests every potential deck against a varied and representative range of human-made decks.
In these preliminary experiments, the approach has proven able to create quite effective decks, a promising result that proves that, even in this challenging environment, evolutionary algorithms can find good solutions.

[Paper] My life as a sim: evolving unique and engaging life stories using virtual worlds

Our latest publication My life as a sim: evolving unique and engaging life stories using virtual worlds, using our framework MADE (created by @rubenhek), has been published in the ALIFE 2014 conference. The abstract:

Stories are not only painfully weaved by crafty writers in the solitude of their studios; they also have to be produced massively for non-player characters in the video game industry or tailored to particular tastes in personalized stories. However, the creation of fictional stories is a very complex task that usually implies a creative process where the author has to combine characters, conflicts and backstories to create an engaging narrative. This work describes a general methodology to generate cohesive and coherent backstories where desired archetypes (universally accepted literary symbols) can emerge in complex stochastic systems. This methodology supports the modeling and parametrization of the agents, the environment where they will live and the desired literary setting. The use of a Genetic Algorithm (GA) is proposed to establish the parameter configuration that will lead to backstories that best fit the setting. Information extracted from a simulation can then be used to create the literary work. To demonstrate the adequacy of the methodology, we perform an implementation using a specific multi-agent system and evaluate the results, testing with three different literary settings.

Check out the presentation by @jjmerelo at http://jj.github.io/alife14-made/#/home. You can download the proceedings of the conference (CC license), or download the paper draft.

More information is available on the project page.