The Santa Fe ant problem is a well-known model problem that has been examined extensively over the past two decades and continues to be actively investigated. However, this problem is considered "difficult" because evolutionary computing technologies cannot solve it more efficiently than random search.

It is a well-studied genetic programming issue introduced by Koza in his landmark book. This problem and the more computationally challenging Los Altos Trail problem appeared in Koza's book.

Overview

The structure of programmes has been determined via fixed-length schema analysis to be "very misleading." The artificial ant problem is frequently used as a benchmark for genetic programming. The problem is to build a programme capable of successfully guiding an artificial ant along a winding path on a 3232 toroidal grid. Move right, and we can use left in the programme to make the ant advance one square, turn to the left, and turn to the right. Each of these procedures requires a single unit of time. If FoodAhead(), the sensing function looks into the square the ant is currently facing and executes one of its two parameters based on whether or not it contains food. 

The artificial insect must follow the "Santa Fe Trail," which consists of 144 squares and 21 turns. It consists of 89 non-uniformly distributed meal units. When an ant visits a square containing food, it consumes the food in that square. The Santa Fe Ant problem is often an illustrative example, among other issues. Commonly, it is used to demonstrate the viability of meta-heuristics or innovative approaches and as a performance benchmark. For example, the problem has been employed in studies of grammatical evolution to evaluate the impact of genotypic diversity, crossover kinds, genome length, degeneration, and wrapping on the percentage of invalid individuals and cumulative frequency of success.

Challenges

Sometimes, among various other challenges, the Santa Fe ant problem is utilized as a representative issue. Typical applications for this purpose include performance benchmarking and demonstrating the efficacy of new techniques or metaheuristics. For example, Chellapilla used the case to show how powerful tree mutations may be in GP. The issue was one of several operated by Fonlupt and Robilliard to illustrate the usefulness of Linear Differential Evolutionary Programming. The case has also demonstrated the effectiveness of Constituent Grammatical Evolution, Analytic Programming, and Cartesian Genetic Programming. Finally, the issue has been utilized to research the negative slope coefficient and generalization capacity.

Evolution

The Santa Fe artificial ant path has also been extensively examined utilizing additional metaheuristics. In Grammatical Evolution, the problem has been used to explore the impact of crossover types (and, to a lesser extent, mutation rates), degeneration and wrapping on the fraction of invalid individuals, genome length, genotypic diversity, and cumulative frequency of success. With and without degeneracy, the effect of changes to standard grammar and bias caused by the addition of grammar-defined introns is studied. Finally, this study examines the impact of grammar size and complexity on performance.

Conclusion

The Santa Fe Ant model issue has been extensively utilized in the past two decades to research, test, and evaluate evolutionary computing systems and approaches. However, there needs to be literature on its programme structures that are systematically employed for fitness enhancement, their geometries, or their dynamics during optimization.

In addition, the arrangement of food pellets in the Santa Fe Trail problem has become a benchmark for assessing the performance of various genetic programming techniques and solutions. Furthermore, using the NetLogo application is one approach for programming and testing methods for the Santa Fe Trail problem. Finally, at least one student has constructed a Lego robotic to address the issue.