## Description Instances for the pedestrian evacuation problem. Each datafile represents an indoor space from which evacuation is attempted. This space is a rectangular area of width $w$ and height $h$ representing the floor plan of an enclosed space and therefore all its boundaries are assumed to be blocked/non-traversable, except in specific locations which will be denoted as _accesses_. Inside the area there may be a number of _obstacles_ denoting non-traversable regions (representing real-world objects such as walls or furniture). ## Format Each environment is described in JSON format. The structure of each JSON object is described in the `environment-schema.json` file in the root folder. ## Contents Each dataset in in a separate folder. * `random`. Instances with randomly distributed obstacles and no pre-defined accesses. The files are named as `base-[low|mid|high]-density-X.json`, where the infix denotes the number $n$ of obstacles ($n\in$ {$20,\dots, 30$}, $n\in$ {$50,\dots, 75$}, and $n \in ${$100,\dots, 150$} respectively), and the suffix is an instance index. In all cases, the width and height are picked from [40, 50] and [20, 30] respectively. These instances have been used in [1]. * `office`. Instances generated with an L-attributed grammar to resemble office floors: open areas, conference tables, cubicles with furniture (desks, chairs, cabinets). The files are named as `office-X.json`, where the suffix is an instance index. * `supermarket`. Instances generated with an L-attributed grammar to resemble supermarket and shopping centers: display stands in sections, aisles, checkout counters, food courts, small shops. The files are named as `supermarket-X.json`, where the suffix is an instance index. ## References 1. Carlos Cotta, José E. Gallardo, "Evolutionary Algorithms for Optimizing Emergency Exit Placement in Indoor Environments", _Applications of Evolutionary Computation – 27th International Conference, EvoApplications 2024_, Stephen Smith, João Correia and Christian Cintrano (eds.), Lecture Notes in Computer Science, Springer, Berlin, 2024