Weighted A Star

A light-weight, reusable, and efficient C++17 implementation of Weighted A* planning algorithm

Features

• Can be easily added to your project
• Uses modern C++ optimizations
• Efficient and Light-Weight

How to use it in your project

• The implementation uses C++ templates, so you can have any data-structure that represents a state in the planning problem. The only requirement is that you also provide a hash function for the same data-structure if it is not available already in C++. Initialize the planner with custom data-structure that represents state stateT and action actionT, optionally along with the custom hash function for the state as such WAstar<stateT, actionT, stateHasher>.
• You just have to provide the planner with three helper functions based on your domain
  • double getH(stateT s, stateT goal): Your implementation of any heuristic function for your domain which takes in a state s and finds the heuristic value to the goal. The return type of double is fixed in my implementation as it works will every case.
  • getSuccessors(stateT s): A function which my planner will call to get the successors of any state s. This allows you to have implicit, explicit or latice-based graphs etc. It returns a more complicated data-structure, so please refer to the code.
  • bool satisfiesGoal(stateT s, stateT goal): Sometimes for your planning problem the goal may be under-represented. Thus, my planner will allow you to specify a custom search termination condition which checks whether a state s satisfies the goal condition given by the state goal.

Installation

• Just include the main header: #include "lib_planner.h"
• Make sure to build with -std=c++17 flag and -O3 would also help.

Feature Tracker

• Add demo planning problem
• Support for a goal region instead of a single goal state

mpdmanash (at) cmu.edu