Skip to content

technicalReport.md

Latest commit

 

History

History
348 lines (264 loc) · 16.2 KB

technicalReport.md

File metadata and controls

348 lines (264 loc) · 16.2 KB

Sucky Robot Technical Report

Overview

The Sucky robot is an advanced autonomous cleaning platform with three main operational capabilities: teleoperation, mapping, and autonomous navigation. This technical report details the system architecture, implementation details, and operational modes of the robot.

1. Teleoperation Mode

The teleoperation mode enables manual control of the robot through a PS4 controller interface.

Launch Configuration

Drive System

  • Control Framework: ROS2 Control with differential drive configuration
  • Hardware Interface: RoboClaw Hardware Interface (Apache 2.0 License)
  • Motor Controllers: RoboClaw motor controller for precise differential drive control

Custom Hardware Integration

The robot utilizes an Arduino microcontroller for additional functionality beyond basic locomotion:

Hardware Components Controlled:

  • Vacuum System: Two independent cyclone units controlled by Electronic Speed Controllers (ESCs)
  • Door Control: Two PWM servo motors for automated door operation
  • Shaker Motor: Relay-controlled motor for debris agitation
  • Airlock Feeder: Relay-controlled feeder mechanism

Communication

  • Interface: USB serial communication between Arduino and Jetson Orin Nano
  • Protocol: Custom bidirectional communication for state information and command execution

Software Architecture

Custom ROS2 Nodes:

  1. arduino_controller.py

    • Bridges Arduino hardware interfaces to ROS topics
    • Provides high-level control abstraction for peripheral systems
    • Manages state information and component interfacing

  2. sucky_joy.py

    • Extension of the standard ROS joy node
    • Interprets joystick button inputs for Arduino-controlled components
    • Provides intuitive control mapping for complex operations

Configuration Files:

Development Tools:

  • arduino_serial.py: Standalone serial terminal for Arduino functionality testing and hardware debugging

Arduino Firmware:

  • Location: sucky_arduino - PlatformIO project containing custom firmware for peripheral control

2. Mapping Mode

The mapping mode combines teleoperation capabilities with advanced SLAM (Simultaneous Localization and Mapping) functionality to create detailed environment maps.

Launch Configuration

Visual SLAM Implementation

The robot employs NVIDIA Isaac ROS Visual SLAM for optimized odometry and SLAM processing:

Key Features:

  • Hardware Acceleration: Leverages Jetson Orin Nano GPU for optimal performance
  • Sensor Integration: Intel RealSense camera providing stereo IR image pairs
  • Transform Publishing: Generates base_footprint → odom transform
  • Map Conflict Resolution: VSLAM map → odom transform disabled to prevent conflicts with SLAM Toolbox

Performance Considerations:

  • Current Limitation: Single RealSense camera configuration
  • Recommended Upgrade: Dual StereoLabs Zed cameras for enhanced VSLAM robustness
  • Hardware Bottleneck: Jetson Orin Nano performance constraints identified

SLAM Toolbox Integration

Map Generation: SLAM Toolbox provides 2D occupancy grid mapping:

Sensor Configuration:

Mapping Performance:

  • Current Status: Functional but with identified improvement areas
  • Primary Bottleneck: Jetson Orin Nano computational limitations
  • Recommended Improvements:
    • 360-degree LiDAR for enhanced localization
    • Improved odometry accuracy
    • Highly accurate URDF model refinement

Alternative Solutions Evaluated

During development, several alternative mapping approaches were investigated:

Isaac ROS Mapping:

  • Visual mapping solution using Isaac ROS framework
  • Development Status: Configuration challenges prevented successful integration with custom platform

3. Navigation Mode

The autonomous navigation mode represents the most advanced operational capability, combining multiple perception and planning systems for fully autonomous operation.

Launch Configuration

Localization System

AMCL Implementation

Adaptive Monte Carlo Localization (AMCL) provides robot localization on pre-generated maps:

  • Current Status: Functional implementation with identified optimization opportunities
  • Computational Impact: High processing requirements
  • Environmental Sensitivity:
    • Susceptible to mapping errors
    • Affected by dynamic environment changes
    • Prone to dust obscuration of LiDAR sensors

Alternative Localization Solutions Evaluated:

  1. Isaac ROS Map Localization:

    • GPU-accelerated lidar localization solution
    • Development Status: Testing revealed accuracy issues; requires further investigation

  2. cuVGL (CUDA Visual-Geometric Localization):

    • Hardware-accelerated visual localization solution
    • Dependency: Requires maps generated using Isaac ROS Mapping visual mapping capability

Navigation Stack

Nav2 Configuration

ROS 2 Navigation Stack (Nav2) handles path planning and execution:

  • Configuration: sucky_nav2.yaml
  • Modification Level: Largely unmodified from standard Nav2 behavior
  • Odometry Source: Isaac ROS Visual SLAM (primary and only current source)

Odometry Fusion Experiments

Extended Kalman Filter (EKF) integration attempted:

  • Components: VSLAM odometry + IMU data + ros2_control
  • Outcome: Resulted in noisy transform; reverted to VSLAM-only configuration

3D Perception and Obstacle Avoidance

NVBLOX Integration

NVIDIA Isaac ROS NVBLOX provides real-time 3D environment reconstruction:

Core Functionality:
  • Input: RealSense stereo IR images
  • Processing: GPU-accelerated 3D voxel mapping
  • Output: Height-filtered 2D costmap projection for Nav2 integration
Costmap Configuration:
  • Local Costmap: NVBLOX serves as primary obstacle detection layer
  • Global Costmap: NVBLOX combined with:
    • SLAM-generated static map
    • Inflation layer
    • Traditional obstacle layer
Performance Characteristics:
  • Range: Local area mapping optimized for immediate navigation needs
  • LiDAR Limitation: Excluded from obstacle avoidance due to dust sensitivity
  • Filter Solution: Particle filter implementation to mitigate false positive detections
Future Enhancement Opportunities:
  • Hardware: Enhanced camera configuration and increased computational power
  • Software: Implementation of people segmentation and additional NVBLOX operational modes

Waypoint Navigation System

The system enables dynamic waypoint navigation with comprehensive obstacle avoidance:

Navigation Interfaces:

  • Foxglove Studio: Visual waypoint selection and monitoring
  • ROS Topic: Direct /goal_pose topic for programmatic control

Capabilities:

  • 3D Obstacle Detection: Real-time static and dynamic obstacle avoidance
  • Path Planning: Intelligent routing around detected obstacles
  • Goal Management: Robust handling of unreachable waypoints

4. Coverage Planning System

Sucky Coverage Node

sucky_coverage.py - Intelligent coverage path planning system:

Functionality:

  • Input: Four rectangular boundary coordinates
  • Output: Array of waypoints implementing lawnmower coverage pattern
  • Algorithm: Systematic area coverage with configurable parameters

Configuration Parameters:

  • Robot Footprint: Accounts for physical robot dimensions
  • Waypoint Density: Configurable spacing between waypoints
  • Edge Margin: Safety buffer from boundaries
  • Configuration File: coverage_config.yaml

Future Development Requirements:

  1. Occupancy Integration:

    • Analyze global map occupancy values during waypoint generation
    • Eliminate waypoints in inaccessible locations
    • Reduce failed navigation attempts to unreachable goals

  2. Nav2 Integration:

    • Utilize Nav2-published robot footprint instead of independent configuration
    • Ensure consistency between navigation and coverage planning systems

5. Mission Management System

High-Level Autonomous Control

The autonomous operation requires sophisticated mission management beyond basic navigation capabilities.

Sucky Missions Node

sucky_missions.py - Mission configuration and execution system:

Supported Mission Types:
  1. Cleaning Tasks: Comprehensive area cleaning operations
  2. Navigate to Point: Targeted navigation missions
  3. Future Extensions (planned):
    • Dust ejection operations
    • Charging dock navigation
Mission Interface Options:
  • Terminal Interface: mission_terminal.py
  • ROS Topic Interface: Direct topic-based mission commands
  • Future State Machine: Planned autonomous decision-making logic

Mission Configuration

  • Mission Definitions: sucky_missions.yaml
  • Coordinate Collection: Manual teleoperation with /amcl_pose recording
  • Future Enhancement: GUI-based mission definition with interactive map interface

System Monitoring

Battery Management

battery_monitor.py - Power system monitoring:

  • Function: Battery voltage monitoring and reporting
  • Development Status: Requires additional work to prevent serial conflicts with RoboClaw and ros2_control
  • Integration: Battery state information for charging mission triggers

Dust Chamber Monitoring

Hardware Integration: Break beam sensor via Arduino controller

  • Function: Empty/full state detection
  • Output: Published ROS topic for system state awareness
  • Integration: Enables automated dust ejection missions

Modular Mission Architecture

Cleaning Missions

  • Design Philosophy: Modular and customizable for deployment flexibility
  • Configuration: Manual coordinate collection via teleoperation
  • Coordinate Source: Recording /amcl_pose during manual exploration

Future GUI Development

Planned Enhancement: Interactive mission definition system

  • Map Display: Visual representation of global map
  • Point-and-Click Interface: Intuitive rectangular area selection
  • Automatic Generation: Streamlined mission definition workflow

Industrial Integration

PLC Communication Capability

Future Development: Industrial automation integration

  • Library: pylogix for PLC communication
  • Applications:
    • Secure area access control
    • Integration with existing automation systems
    • Coordinated industrial cleaning operations

6. System Architecture Summary

Package Organization

The Sucky robot system is organized into several key ROS2 packages:

Core Packages:

  • sucky - Main robot package containing launch files, configurations, and custom nodes
  • sucky_arduino - Arduino firmware for peripheral control systems
  • roboclaw_hardware_interface - Hardware interface for RoboClaw motor controllers
  • roboclaw_serial - Serial communication package for RoboClaw integration

External Dependencies:

Key Configuration Files:

File Location Purpose
joystick.yaml Control Config Joystick button mapping
sucky_controllers.yaml Control Config ROS2 Control configuration
sucky_nav2.yaml Navigation Config Nav2 stack parameters
coverage_config.yaml Navigation Config Coverage planning parameters
sucky_missions.yaml Navigation Config Mission definitions

Launch File Architecture:

Launch File Purpose Key Components
teleop.launch.py Manual Control Joy control, Arduino interface
mapping.launch.py SLAM Mapping VSLAM, SLAM Toolbox, sensors
navigation.launch.py Autonomous Nav AMCL, Nav2, NVBLOX
sucky_bringup.launch.py Unified Launch Mode selection interface

7. Future Development Roadmap

Immediate Improvements (Short-term):

  1. Enhanced Localization: Implement Isaac ROS Map Localization to replace AMCL
  2. Coverage Optimization: Integrate occupancy map data into waypoint generation
  3. Battery Integration: Complete battery monitoring system without serial conflicts
  4. GUI Development: Create interactive mission definition interface

System Upgrades (Medium-term):

  1. Hardware Enhancements:

    • Upgrade to dual StereoLabs Zed cameras
    • Implement 360-degree LiDAR system
    • Enhanced computational platform (higher-end Jetson)

  2. Software Improvements:

    • Advanced sensor fusion with EKF integration
    • People segmentation in NVBLOX
    • State machine for autonomous decision making

Advanced Features (Long-term):

  1. Industrial Integration:

    • Complete PLC communication implementation
    • Secure area access protocols
    • Industrial automation coordination

  2. Advanced Perception:

    • Multiple NVBLOX operational modes
    • Enhanced dust detection and avoidance
    • Improved robustness in dynamic environments

8. References and External Resources

Key Technologies:

Hardware Integration:

Development Tools: