This project provides a Prolog implementation of the Graphplan algorithm proposed in:
Blum, Avrim L., and Merrick L. Furst. "Fast planning through planning graph analysis." Artificial intelligence 90.1-2 (1997): 281-300.
Furthermore, an explanation of the plans obtained is provided in order to visualize the intermediate states of the planning process.
The domains are defined according to the STRIPS representation, in this case it is also possible to specify the application cost of each action.
Two domains related to rocket scenario and public transfers are already defined.
Each domain is separated in a file containing the operations and in another file containing the initial state and the goal to be reached.
The planner has been tested with SWI-Prolog 7.6.4 and YAProlog 6.2.2.
First of all, you need to load the main module:
consult('planner_start.pl').
Then calling the predicate start_planner will display a Text GUI to interact with the planner:
start_planner.
-----------------------------------------------------------------------
AUTOMATIC PLANNING MODULE
-----------------------------------------------------------------------
1. Load domain
2. Load initial & goal state
3. Plan
4. Explanation
0. Exit
|:
You need to specify the path of the domain and the states for the planning:
|: 1.
Insert path of domain: |: 'domain/rocket'.
Domain loaded correctly
|: 2.
Insert path of states: |: 'domain/rocket_states'.
Initial state: at(a,london) at(b,london) at(rocket1,london) has_fuel(rocket1)
Goal: at(a,paris) at(b,paris)
Then you can start the planning:
|: 3.
Plan found. Id: 1
Step 1: load(rocket1,london,a) load(rocket1,london,b)
Step 2: move(rocket1,london,paris)
Step 3: unload(rocket1,paris,a) unload(rocket1,paris,b)
Cost: 5
Explanations are provided by specifying the id of a returned plan:
|: 4.
Plan id: |: 1.
In the state: has_fuel(rocket1) at(a,london) at(b,london) at(rocket1,london)
I applied the actions: load(rocket1,london,a) load(rocket1,london,b)
to obtain the new state: has_fuel(rocket1) at(rocket1,london) in(a,rocket1) in(b,rocket1)
In the state: has_fuel(rocket1) at(rocket1,london) in(a,rocket1) in(b,rocket1)
I applied the action: move(rocket1,london,paris)
to obtain the new state: at(rocket1,paris) in(a,rocket1) in(b,rocket1)
In the state: at(rocket1,paris) in(a,rocket1) in(b,rocket1)
I applied the actions: unload(rocket1,paris,a) unload(rocket1,paris,b)
to obtain the new state: at(a,paris) at(b,paris)
Alternatively, the planner can be used without the Text GUI by:
- consulting a domain
- calling the predicate plan/3 specifying initial state and goal and returns plans in PlanSet
- calling the predicate plan_explanation/1 specifying the id of the plan to be explained.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent