Skip to content

Latest commit

 

History

History
57 lines (43 loc) · 2.19 KB

README.md

File metadata and controls

57 lines (43 loc) · 2.19 KB
Raw

Traffic Management

A simple programme to manage traffic in a simplified model of a city, based on planning (PDDL) and utility-driven search algorithm (see the task specifications file: 'assessment_CS5060_ASMNT_C_2019_2020.pdf').

Installation and running

Special dependencies

The package requires package PyDDL to be installed. It can be found at: https://github.com/garydoranjr/pyddl.

It can be installed by executing the following command:

C:\Users\xyz> pip install -e git+https://github.com/garydoranjr/pyddl.git#egg=pyddl

Detailed/alternative installation instructions are available at the project home page.

Installation

To install the program, please run the setup.py script:

C:\Users\xyz> cd path_to_the_project\traffic_management
C:\path_to_the_project\traffic_management> python setup.py install

Execution

The main planner script is: 'traffic_management/traffic_planner/traffic_lights_system.py' After navigating to the 'traffic_management/traffic_planner/' directory, it can be executed by:

C:\...\traffic_management> cd traffic_planner
C:\...\traffic_management\traffic_planner> python traffic_lights_system.py

To learn about the available options to adjust the planning problem, use:

C:\...\traffic_management\traffic_planner> python traffic_lights_system.py --help

Disclaimer

Code in package custom_pyddl is a customised version of PyDDL.

The aforementioned customisation includes:

  • for PDDL-like domain and problem definition:
    • subclassing of certain classes (Action, Problem, State) - with the main purpose of adjusting the way of handling numerical effects of actions (see the respective docstrings)
    • rewriting some of the classes (GroundedAction), whose original structure was not suitable for subclassing; this heavily uses the original code, but also adds/alters some functionalities
  • for the planner:
    • rewriting the planner in a form of a class
    • borrowing a considerable part of the original planner to maintain its compatibility with the domain and problem definition (usage, syntax, etc.)
    • replacing the A* search technique with a utility-driven hill-climbing algorithm