Ealain: A Camera Simulation Tool to Generate Instances for Multiple Classes of Optimisation Problem
Quentin Renau, Johann Dréo, Emma Hart · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2024
Artificial benchmark datasets are common in both numerical and discrete optimisation domains. Existing benchmarks cover a broad range of classes of optimisation, but as a general rule have limited value due to their poor resemblance to real-world problems, and generally lack the ability to generate arbitrary numbers of instances. In this paper, we introduce Ealain, an instance-generator that creates instances of optimisation problems which require placement of a number of cameras in a domain --- this has many real-world analogies for example in environmental monitoring or providing security in a building. The software provides two types of camera-model and can be used to generate an infinite number of instances of black-box, real-world-like optimisation problems which can be single-objective, multi-objective, multi-fidelity, or constrained. The software is also flexible in that it also permits a range of different objective functions to be defined. Furthermore, generated instances can be solved using either a numerical or discrete encoding of solutions. The C++ library targets fast computation and can be easily plugged into a solver of choice. We summarise the key features of the Ealain software and provide some examples of the type of instances that can be generated for different classes of optimisation problems.