This repo includes the implementation of GP4 Algorithm described in the paper GP4: Gaussian Process Proactive Path Planning for the Stochastic On Time Arrival Problem, and the benchmarks, namely DOT, Pruning, PLM, ILP and OS-MIP used for comparison.

The network data used for conducting simulations is available at Networks.zip. Please unzip the downloaded file into the same directory of the source code. The dataset includes 8 networks/time slots. Each of them consists of two files:

• .csv, nodes and links of the network, and mean cost of each link.
• .npy, covariance matrix used in experiments.

• Python 3.6+
• NumPy
• SciPy
• Pandas
• NetworkX
• gurobipy (a license might be needed)

The source code includes the following files:

• main.py, sample codes for testing GP4 and benchmarks on simple network described in the paper (Subsection V-C1). Parameters are also specified here.
• GP4.py, implementation of GP4, Log-GP4 and Bi-GP4 Algorithms.
• benchmark.py, our implementation of PLM, ILP, OS-MIP, DOT, and Pruning algorithm in GP-regulated stationary environment.
• evaluation.py, functions that evaluate the on-time-arrival probability of a path.
• func.py, tool functions used throughout the project.