RadarScheduling: A Benchmark Dataset for Radar Scheduling Problems
- Description RadarScheduling is a benchmark dataset designed for the implementation and evaluation of mathematical models and algorithms in radar scheduling. The dataset provides realistic, systematically generated problem instances that simulate different target behaviors under varying operational conditions. By following structured data generation principles, this dataset ensures that the scenarios closely represent real-world radar environments, making it suitable for both academic research and practical algorithm development.
The data instances were generated using MATLAB due to its advanced simulation capabilities, enabling accurate modeling of target motion behavior. Additionally, one real-world instance was used as the basis for creating small-, medium-, and large-scale scenarios.
- Dataset Structure
The dataset is organized into 19 folders. Each folder corresponds to a specific number of initial targets detected by the radar. For example:
- "Folder 5" contains instances where the radar initially detects 5 targets.
- "Folder 20" contains instances starting with 20 targets.
- ...
- "Folder N" corresponds to N initial targets.
Each folder includes 30 problem instances. While all instances within a folder start with the same number of targets, they differ in other parameters, such as target behaviors across time horizons.
- Data Format
Each problem instance is represented by a text file containing 30 consecutive time intervals.
For each interval:- The first number indicates the number of targets present in that interval
- For each target, three values follow:
- Range (0 to 360 km)
- Direction (+1 for approaching targets, −1 for receding targets)
- Velocity (60 to 500 m/s)
These values are listed sequentially for all targets in that interval.
In summary, each file represents the evolution of targets over 30 consecutive horizons (intervals), with varying target behaviors in each horizon.