GOAP-AI

Generate a plan using the GOAP algorithm.

Usage

Define the agent's world view, goals, and actions in a config.yml file.

Agent's world view:

state:
  properties:
    energy: 50
    health: 20
    consciousness: 10
    raw_meat: 0
    cooked_meat: 0
    wood: 0
    fire: 0

Agent's goals:

goals:
  - weight: 1
    property: "energy"
    target: 100
  - weight: 1
    property: "health"
    target: 30
  - weight: 1
    property: "consciousness"
    target: 10
  - weight: 1
    property: "fire"
    target: 1

Agent's action set:

actions:
  - label: "gather_wood"
    duration: 2
    deltas:
      energy: -10
      consciousness: -10
      wood: 10
  - label: "build_fire"
    duration: 1
    deltas:
      energy: -5
      consciousness: -5
      wood: -10
      fire: 10
  - label: "hunt"
    duration: 2
    deltas:
      energy: -10
      health: -2
      consciousness: -40
      raw_meat: 3
  - label: "cook_meat"
    duration: 1
    deltas:
      energy: -5
      health: 0
      consciousness: -5
      raw_meat: -1
      cooked_meat: 1
      fire: -1
  - label: "eat_raw_meat"
    duration: 1
    deltas:
      energy: 5
      health: 2
      consciousness: -10
      raw_meat: -1
  - label: "eat_cooked_meat"
    duration: 1
    deltas:
      energy: 30
      health: 5
      consciousness: -10
      cooked_meat: -1
  - label: "exercise"
    duration: 2
    deltas:
      energy: -10
      health: 5
      consciousness: -20
  - label: "sleep"
    duration: 8
    deltas:
      energy: -10
      health: 5
      consciousness: 80
  - label: "sleep_with_fire"
    duration: 8
    deltas:
      health: 10
      consciousness: 100
      fire: -1
  - label: "wait"
    duration: 1
    deltas:
      energy: -2
      consciousness: -1

Then run the program with the following command:

cargo run --release

You'll see the agent's plan to achieve its goals:

# Solution:
Plan found with score 0.00 and time 31

# Plan:
Step 1) 	d=1.77 t=8	sleep
Step 2) 	d=1.87 t=10	gather_wood
Step 3) 	d=0.92 t=11	build_fire
Step 4) 	d=0.75 t=19	sleep_with_fire
Step 5) 	d=0.85 t=21	hunt
Step 6) 	d=0.90 t=22	cook_meat
Step 7) 	d=0.60 t=23	eat_cooked_meat
Step 8) 	d=0.65 t=24	cook_meat
Step 9) 	d=0.35 t=25	eat_cooked_meat
Step 10) 	d=0.40 t=26	cook_meat
Step 11) 	d=0.10 t=27	eat_cooked_meat
Step 12) 	d=0.20 t=29	hunt
Step 13) 	d=0.25 t=30	cook_meat
Step 14) 	d=0.00 t=31	eat_cooked_meat

# Final state:
consciousness: 35
wood: 0
energy: 105
fire: 5
cooked_meat: 0
raw_meat: 2
health: 51

API

The world state is a list of key-value pairs. The goals are a list of key-value pairs with a target value and a weight. The actions are a list of key-value pairs with a label, duration, and a list of key-value pairs which represent the state changes when the action is executed.

State -> GOAP -> Plan