Aller au contenu principal

Behavior Tree Runtime Debugger

The Behavior Tree Debugger is a powerful real-time visualization tool for inspecting and debugging AI behavior during gameplay. It runs in a separate window and provides complete visibility into decision-making processes.

Overview​

The debugger provides:

  • Real-time visualization of tree execution
  • Entity filtering to focus on specific AI
  • Execution history to track decision patterns
  • Blackboard inspection to view runtime state
  • Performance metrics for optimization

Opening the Debugger​

During Gameplay​

Method 1: Keyboard Shortcut

  • Press Ctrl+Shift+B (Windows/Linux)
  • Press Cmd+Shift+B (macOS)

Method 2: Menu

  1. Click Debug in the main menu
  2. Select Behavior Tree Debugger

Result: A separate window opens with the debugger interface.

Standalone Mode​

You can also launch the debugger before starting the game:

  1. Start Olympe Engine
  2. Open debugger (it remains hidden)
  3. Launch level
  4. Debugger automatically populates with entities

Interface Layout​

The debugger window consists of three main panels:

┌─────────────────────────────────────────────────┐
│  [Entity List]  │  [Node Graph]  │ [Inspector] │
│                 │                 │             │
│  - Guard 1      │   ┌─Selector─┐ │ Runtime Info│
│  - Guard 2      │   │  Combat  │ │ Blackboard  │
│  - Zombie 1     │   └──────────┘ │ Exec Log    │
│                 │                 │             │
└─────────────────────────────────────────────────┘

1. Entity List Panel (Left)​

Displays all entities with behavior trees currently active.

Features:

  • Entity name and ID
  • Tree name
  • Current execution status
  • Last update time
  • Filter and sort controls

Interactions:

  • Click entity to select and view its tree
  • Double-click to center in game view
  • Right-click for context menu

2. Node Graph Panel (Center)​

Visual representation of the behavior tree for the selected entity.

Features:

  • Node hierarchy display
  • Execution status highlighting
  • Connection visualization
  • Pan and zoom controls
  • Layout options (vertical/horizontal)

Status Indicators:

  • 🔵 Blue highlight - Currently Running
  • ✅ Green highlight - Succeeded (last execution)
  • ❌ Red highlight - Failed (last execution)
  • ⚪ Gray - Idle (not executed recently)

3. Inspector Panel (Right)​

Detailed information about the selected entity's AI state.

Sections:

  • Runtime Info - Current node, status, execution time
  • Blackboard - All runtime variables
  • Execution Log - Recent decision history
  • Performance - CPU time, evaluation count

Using the Entity List​

Viewing Entities​

The list shows all entities with AIBehaviorTree_data components.

Columns:

  • Name - Entity name (e.g., "Guard", "Zombie")
  • Tree - Behavior tree name
  • Status - Current execution status
  • Mode - AI mode (Patrol, Combat, etc.)

Filtering Entities​

Text Filter:

  1. Click the filter box (or press Ctrl+F)
  2. Type entity name or tree name
  3. List updates in real-time

Filter Options:

  • ☑️ Active Only - Show only entities currently executing
  • ☑️ Has Target - Show only entities with targets

Example:

  • Type Guard → Shows all guards
  • Type Patrol → Shows all using patrol tree
  • Enable "Active Only" → Shows only moving/attacking entities

Sorting​

Click column headers to sort:

  • Name - Alphabetical order
  • Tree Name - Group by behavior tree
  • Last Update - Most recently active first
  • AI Mode - Group by current behavior mode

Node Graph Visualization​

Reading the Graph​

Node Layout:

        ┌─────────────┐
        │  Selector   │ Root node (top)
        └──────┬──────┘
               │
       ┌───────┴───────┐
       │               │
  ┌────┴────┐    ┌────┴────┐
  │ Combat  │    │ Patrol  │ Children (below)
  └─────────┘    └─────────┘

Vertical Layout (Default):

  • Root at top
  • Children below
  • Execution flows downward

Horizontal Layout:

  • Root at left
  • Children to the right
  • Execution flows left-to-right

Node Colors by Type​

Node TypeColor
SelectorPurple
SequenceBlue
ConditionYellow
ActionGreen
InverterOrange
RepeaterTeal

Execution Highlights​

Nodes show their current execution state:

Running (Blue Pulse):

┌─────────────┐
│ Move to Goal│ ← Animated blue glow
└─────────────┘

This node is currently executing.

Success (Green Flash):

┌─────────────┐
│ Target Check│ ← Green tint
└─────────────┘

This node succeeded in the last evaluation.

Failure (Red Flash):

┌─────────────┐
│ Can Attack? │ ← Red tint
└─────────────┘

This node failed in the last evaluation.

Execution Token​

A token (small circle) animates through the tree showing the current execution path:

Selector
 ├─ Combat ➜ ● ← Token here (failed)
 └─ Patrol ● ← Token moves here

Camera Controls​

Zoom​

Mouse Wheel:

  • Scroll up → Zoom in
  • Scroll down → Zoom out
  • Range: 30% to 300%

Keyboard:

  • + or = → Zoom in (120%)
  • - → Zoom out (80%)
  • 0 → Reset to 100%

Current zoom: Displayed in top-right corner (e.g., "Zoom: 150%")

Pan​

Middle Mouse Button:

  • Click and drag to pan view

Space + Drag:

  • Hold Space, then drag with left mouse

Keyboard (when node selected):

  • Arrow keys → Pan in direction

Framing​

Fit to View:

  • Press Home or click Fit to View button
  • Centers entire tree in viewport

Center on Selected:

  • Press F or double-click entity in list
  • Centers on selected entity's tree

Reset Camera:

  • Click Reset Camera button
  • Resets zoom to 100% and centers tree

Layout Options​

Switching Layouts​

Vertical (Top-to-Bottom):

  • Click Vertical button
  • Or press Ctrl+1
  • Best for: Deep trees, many levels

Horizontal (Left-to-Right):

  • Click Horizontal button
  • Or press Ctrl+2
  • Best for: Wide trees, many children

Spacing Controls​

Adjust Node Spacing:

  1. Find spacing sliders in toolbar
  2. Horizontal Spacing: 150-500px (default: 250px)
  3. Vertical Spacing: 100-400px (default: 180px)
  4. Drag sliders to adjust

Reset to Defaults:

  • Click Reset Spacing button
  • Restores 250×180px spacing

Auto Layout:

  • Click Auto Layout button
  • Automatically arranges nodes for best visibility

Inspector Panel​

Runtime Info Section​

Displays current execution state:

Fields:

  • Current Node: Name and ID of executing node
  • Status: Running/Success/Failure
  • Tree ID: Behavior tree asset ID
  • Tree Name: Loaded tree file name
  • Execution Time: Total time spent in tree this frame (ms)

Example:

Current Node: Move to Goal (#12)
Status: Running
Tree: guard_ai.json (ID: 42)
Execution Time: 0.8ms

Blackboard Section​

Shows all runtime variables stored in AIBlackboard_data.

Common Variables:

VariableTypeDescription
targetEntityEntityIDCurrent target (0 = none)
lastKnownTargetPositionVectorLast seen target location
moveGoalVectorCurrent movement destination
waitTimerfloatTimer for wait actions
heardNoiseboolNoise detection flag
noisePositionVectorLocation of noise
reachedDestinationboolArrived at goal?

Example Display:

Blackboard:
  targetEntity: 12
  moveGoal: (450, 320)
  waitTimer: 0.0
  lastKnownTargetPosition: (500, 300)

Features:

  • Values update in real-time
  • Click to inspect details
  • Editable in debug mode (advanced)

Execution Log​

Chronological list of recent node executions.

Log Entry Format:

[Time] Entity | Node Name | Status

Example Log:

[0.0s] Guard #5 | Guard Decision | Running
[0.0s] Guard #5 | Combat | Failure
[0.1s] Guard #5 | Patrol | Running
[0.1s] Guard #5 | Pick Next Waypoint | Success
[0.2s] Guard #5 | Move to Waypoint | Running
[3.5s] Guard #5 | Move to Waypoint | Success
[3.6s] Guard #5 | Patrol | Success

Features:

  • Auto-scrolls to latest entry
  • Max 100 entries (circular buffer)
  • Color-coded by status
  • Filterable by entity

Controls:

  • Clear Log: Removes all entries
  • Pause: Stops logging new entries
  • Copy: Copy log to clipboard
  • Save: Export log to file

Debugging Workflows​

Workflow 1: Why isn't my AI working?​

Steps:

  1. Open debugger (Ctrl+Shift+B)
  2. Find entity in Entity List
  3. Select entity to view tree
  4. Watch Node Graph:
    • Which nodes are Running?
    • Which nodes are Failing?
  5. Check Inspector:
    • Is targetEntity set correctly?
    • Are movement goals valid?
  6. Review Execution Log:
    • See decision pattern over time
    • Identify stuck nodes

Common Issues:

  • Node always Running: Action never completes (check logic)
  • Node always Failing: Condition never met (check parameters)
  • Tree stuck on one branch: Selector/Sequence not re-evaluating

Workflow 2: Why did the AI make that decision?​

Steps:

  1. Select entity after unexpected behavior
  2. Check Execution Log:
    • See exact sequence of decisions
    • Identify which condition triggered behavior
  3. Review Blackboard:
    • What values caused the condition to pass?
  4. Replay in slow motion:
    • Pause game
    • Step frame-by-frame (N key)
    • Watch decision tree evaluate

Example: "Why did the guard attack instead of patrol?"

Log shows:
[10.2s] Combat | Running
[10.2s] Target Visible? | Success ← Here's why!

Blackboard shows:
  targetEntity: 7 ← Player was detected

Workflow 3: Performance optimization​

Steps:

  1. Enable performance metrics
  2. Sort entities by Last Update
  3. Identify slow evaluating entities
  4. Select entity to view tree
  5. Check Execution Time in Inspector
  6. Identify expensive nodes (>1ms)
  7. Optimize:
    • Add conditions to early-exit
    • Cache expensive calculations in blackboard
    • Simplify tree structure

Performance Tips:

  • Trees should evaluate in <0.5ms
  • Use conditions to skip expensive branches
  • Limit tree depth (2-4 levels ideal)

Advanced Features​

Minimap​

Toggle: Press Ctrl+M or click Minimap button

A small overview of the entire tree appears in corner:

┌─────────┐
│ ┌─┐     │
│ │ │ ←●  │ ← You are here
│ └─┘     │
└─────────┘

Usage:

  • See where you are in large trees
  • Click to jump to location
  • Useful for trees with 20+ nodes

Grid Display​

Toggle: Press Ctrl+G or click Grid button

Displays alignment grid for node positioning:

  • Helps align nodes visually
  • Shows spacing increments
  • Useful for manual layout

Playback Control (Experimental)​

Some debugger builds support playback control:

Features:

  • Pause (Space) - Pause AI execution
  • Step Frame (N) - Advance one frame
  • Step Node (Shift+N) - Execute one node

Note: This feature may not be available in all builds.

Troubleshooting​

Debugger window blank​

Causes:

  • No entities with behavior trees in scene
  • Trees not loaded successfully
  • Debugger not initialized

Solutions:

  • ✅ Spawn entity with AIBehaviorTree_data
  • ✅ Check console for tree load errors
  • ✅ Restart debugger window

Can't see entity in list​

Causes:

  • Entity doesn't have behavior tree
  • Filter is active
  • Entity just spawned (refresh pending)

Solutions:

  • ✅ Clear filter (Escape)
  • ✅ Check entity has AIBehaviorTree_data
  • ✅ Wait for auto-refresh (0.5s interval)

Nodes not highlighting​

Causes:

  • Tree not executing (entity disabled)
  • Debugger out of sync

Solutions:

  • ✅ Check entity is active in game
  • ✅ Close and reopen debugger
  • ✅ Select different entity, then re-select

Keyboard Shortcuts Quick Reference​

ShortcutAction
Ctrl+Shift+BToggle debugger window
Ctrl+FFilter entities
Mouse WheelZoom
+ / -Zoom in/out
0Reset zoom
Middle Mouse DragPan
HomeFit to view
FCenter on selection
Ctrl+1Vertical layout
Ctrl+2Horizontal layout
Ctrl+MToggle minimap
Ctrl+GToggle grid
Up/DownNavigate entity list
EnterCenter on entity tree

Best Practices​

1. Keep Debugger Open During Development​

  • Always have debugger visible when testing AI
  • Catch issues immediately
  • Understand AI decision patterns

2. Use Filters Effectively​

  • Focus on specific entity type
  • Filter by active entities only
  • Reduce visual clutter

3. Monitor Execution Log​

  • Watch for unexpected patterns
  • Identify infinite loops (same node repeating)
  • Track state transitions

4. Check Blackboard Regularly​

  • Verify variables are set correctly
  • Ensure conditions have correct data
  • Debug logic errors early

5. Learn Your Trees' Normal Behavior​

  • Know what successful execution looks like
  • Spot anomalies quickly
  • Understand performance baseline

Next Steps​

Now that you understand the debugger:

Additional Resources​

Pro Tip: Always run the debugger on a second monitor if available. This lets you see both the game and AI visualization simultaneously for maximum insight!