Behavior Executor Node

Loads and executes behavior trees for autonomous robot control.

Overview

The behavior executor node runs behavior trees at a configurable tick rate, providing autonomous control for robots. It supports enable/disable control from the dashboard and advanced interrupt/resume functionality for seamless integration with manual control.

Features

• Behavior Tree Execution: Loads and runs behavior trees from JSON files
• Built-in Actions: Wander, move, timer, sensor_check nodes
• Configurable Tick Rate: Control execution frequency
• Dashboard Control: Enable/disable via dashboard UI
• Smart Interruption: Pause for manual commands, auto-resume when done
• Status Publishing: Real-time execution status and statistics
• Health Monitoring: Reports behavior tree health and tick counts

Configuration

# Behavior Executor Configuration
behavior_name = "patrol"         # Name of behavior to load (from behaviors/)
behaviors_dir = "behaviors"      # Directory containing behavior JSON files
tick_rate_hz = 10.0             # Execution frequency in Hz
max_ticks = 10000               # Optional maximum tick count
log_stats = true                # Log periodic statistics

[topics]
control = "/behavior/control"    # Control topic for enable/disable/interrupt/resume
status = "/behavior/status"      # Status publishing topic

Control Messages

The behavior executor listens to /behavior/control for control commands:

Basic Control (Dashboard)

Enable Behavior Tree

{
  "action": "enable",
  "timestamp": 1234567890
}

Disable Behavior Tree

{
  "action": "disable",
  "timestamp": 1234567890
}

Advanced Control (Node Interruption)

Interrupt with Auto-Resume

{
  "action": "interrupt",
  "source": "llm-command",
  "duration_secs": 30,
  "timestamp": 1234567890
}

When duration_secs is provided, the behavior tree will automatically resume after the timeout.

Manual Resume

{
  "action": "resume",
  "source": "llm-command",
  "timestamp": 1234567890
}

Resumes the behavior tree immediately, cancelling any pending auto-resume.

Status Messages

Published to /behavior/status every second:

{
  "behavior_name": "patrol",
  "is_running": true,
  "tick_count": 12345,
  "timestamp": 1234567890
}

Control Flow

Startup

1. Node starts with behavior execution disabled
2. Waits for enable command from dashboard
3. When enabled, resets behavior tree and starts ticking

Interruption Flow

1. Behavior tree is running (enabled)
2. Receives "interrupt" command from node (e.g., llm-command)
   → Behavior execution pauses
   → If duration_secs provided, schedules auto-resume
3. Manual command executes
4. After timeout OR manual resume:
   → Behavior tree resets and resumes execution

Manual Control

Dashboard:
  - "Enable" button → Starts behavior tree
  - "Disable" button → Stops behavior tree

Nodes:
  - Send "interrupt" → Pauses for user command
  - Send "resume" → Resumes autonomous behavior

Behavior Tree Files

Behavior trees are loaded from JSON files in the behaviors/ directory:

Example: behaviors/patrol.json

{
  "type": "Sequence",
  "children": [
    {
      "type": "wander",
      "params": {
        "linear_speed": 0.3,
        "angular_speed": 0.5,
        "duration_secs": 5.0
      }
    },
    {
      "type": "timer",
      "params": {
        "duration_secs": 2.0
      }
    }
  ]
}

Built-in Action Nodes

Wander

Moves the robot with random velocity:

{
  "type": "wander",
  "params": {
    "linear_speed": 0.3,
    "angular_speed": 0.5,
    "duration_secs": 5.0
  }
}

Move

Moves with fixed velocity:

{
  "type": "move",
  "params": {
    "linear": 0.5,
    "angular": 0.0,
    "duration_secs": 3.0
  }
}

Timer

Waits for a duration:

{
  "type": "timer",
  "params": {
    "duration_secs": 2.0
  }
}

Sensor Check

Checks sensor values:

{
  "type": "sensor_check",
  "params": {
    "sensor_topic": "/sensors/proximity",
    "condition": "greater_than",
    "threshold": 0.5
  }
}

Integration with LLM Command

The behavior executor works seamlessly with the llm-command node:

1. Robot is patrolling (behavior tree enabled)
2. User says "stop" via LLM
3. llm-command sends interrupt with 30s timeout
4. Behavior tree pauses, robot stops
5. After 30 seconds, behavior tree auto-resumes
6. Robot continues patrolling

This provides a natural "take control, then hand back" UX.

Running

The node is launched automatically by mecha10 dev when included in mecha10.json.

To run manually:

cargo run -p mecha10-nodes-behavior-executor

Health Check

The health check monitors:

• Behavior execution status (Success/Failure)
• Tick count
• Last status

If the behavior status is Failure, the health check reports unhealthy.

Dependencies

• mecha10-core: Framework core (Context, Topic, Message)
• mecha10-behavior-runtime: Behavior tree runtime
• tokio: Async runtime
• serde/serde_json: Serialization
• anyhow: Error handling

See Also

• mecha10-behavior-runtime - Behavior tree implementation
• llm-command - Natural language commands with interruption
• Behavior Interrupt System - Reusable interrupt trigger