ReGoap

ReGoap

ReGoap is a generic C# GOAP (Goal Oriented Action Planning) library with Unity and Godot adapters.

It gives you a planner that chooses a goal, builds a valid action sequence to satisfy it, and lets your runtime execute that sequence while reacting to world changes.

What GOAP is

GOAP is a decision-making technique for agents (NPCs, workers, enemies, simulation entities) where behavior is generated from world facts instead of hardcoded behavior trees or giant switch statements.

In GOAP, you define:

• World state facts (example: hasOre = true, chestSwordCount = 3)
• Actions with preconditions and effects (example: SmeltIngot needs ore and produces ingot)
• Goals as desired world states (example: chestSwordCount >= 10)

At runtime, the planner searches for a valid chain of actions that transforms the current world state into the goal state.

Why this is useful:

• You add/modify actions and goals without rewriting all transitions.
• Agents can recover from changed world conditions via replanning.
• The system naturally supports emergent sequences from reusable actions.

What ReGoap does

ReGoap provides:

• A reusable, engine-agnostic planner core (ReGoap/Core, ReGoap/Planner, ReGoap/Utilities)
• Runtime abstractions for agent, action, goal, memory, and sensors
• Engine adapters (Unity and Godot) so you can plug into scene/component workflows
• Debugging support (notably in Godot runtime debugger)

At a high level, each planning cycle is:

1. Agent gathers possible goals.
2. Planner picks a goal (deterministic or weighted-random mode).
3. Planner runs A* search to build a valid action plan.
4. Agent executes actions in order.
5. On success/failure/world change, agent can re-evaluate and replan.

Core concepts

State

A key/value fact container describing world conditions.

Examples:

• enemyVisible = true
• weaponEquipped = false
• chestOreCount = 2

Action

A behavior unit with:

• Preconditions: what must be true to run it
• Effects: what it contributes toward goals
• Cost: planning cost used by search
• Runtime Run logic: your actual gameplay/sim code

Goal

A desired target state plus priority.

The planner chooses among possible goals and attempts to build a valid plan for one.

Memory

The agent's current world knowledge.

This is the source-of-truth state the planner reads when generating plans.

Sensor

A memory updater that writes observed world facts into memory (resource counts, visibility, occupancy, etc).

New/global planner features

Comparator conditions

ReGoap supports comparator-based state matching through ReGoapCondition:

• ReGoapCondition.Equal(value)
• ReGoapCondition.NotEqual(value)
• ReGoapCondition.GreaterOrEqual(value)
• ReGoapCondition.LessOrEqual(value)

This is useful for count-based logic such as inventories, chest resources, cooldown thresholds, and score targets.

Plain values still use exact equality matching for backward compatibility.

Weighted-random goal selection

Goal selection can be deterministic (highest priority first) or weighted-random (priority-biased random).

Settings in ReGoapPlannerSettings:

• UseWeightedRandomGoalSelection (default false)
• WeightedRandomGoalPriorityPower (default 1f)
• WeightedRandomMinimumWeight (default 0.001f)
• WeightedRandomUseDeterministicSeed (default false)
• WeightedRandomSeed (default 0)

Deterministic seed mode is helpful for reproducible tests/replays.

Repository layout

• ReGoap/Core: interfaces, states, action/goal contracts
• ReGoap/Planner: planner, nodes, settings
• ReGoap/Utilities: logging and utility helpers
• ReGoap/Unity: Unity adapter + Unity examples/docs
• ReGoap/Godot: Godot adapter + Godot FSMExample/docs

Engine-specific guides

• Unity guide: ReGoap/Unity/README.md
• Godot guide: ReGoap/Godot/README.md

Typical integration flow

1. Define your world facts in memory/sensors.
2. Implement actions with realistic preconditions/effects/costs.
3. Implement goals with priorities.
4. Attach agent + planner manager.
5. Validate behavior in debugger and with tests.
6. Tune costs/priorities to shape behavior.

Contributing

PRs are welcome. Run relevant build/tests for changed engine/runtime areas before submitting.