AuTom and Jerry

Posted on In

Autonomous Robot for Target Detection and Pursuit

Overview

AuTom and Jerry is an autonomous robotic system inspired by the classic “Push-Button Kitty” episode of Tom and Jerry. The project demonstrates an intelligent mobile robot capable of detecting, tracking, and approaching moving targets in unknown environments while avoiding obstacles.

Key Capabilities:

  • Real-time target detection and tracking
  • Autonomous navigation with obstacle avoidance
  • Dynamic path planning in unknown environments
  • Robust state management for various operational scenarios

System Architecture

Hardware Components

The system is built on a standard MBot platform with the following components:

Core Components:

  • Raspberry Pi (main processing unit)
  • BeagleBone (auxiliary processing)
  • LiDAR sensor (environmental mapping)
  • Two-wheeled differential drive system
  • On-board camera system

Enhanced Components:

  • Two additional 720p cameras with 100° field of view
  • Replaced standard PiCam to increase detection range and coverage

Software Framework

The system integrates several key technologies:

Computer Vision:

  • AprilTag detection for target identification
  • Real-time pose estimation
  • Multi-camera sensor fusion

Navigation & Planning:

  • Simultaneous Localization and Mapping (SLAM)
  • A* pathfinding algorithm
  • PID control for precise motion

State Management:

  • Finite State Machine (FSM) for behavioral control
  • Robust handling of target loss and recovery

Technical Implementation

Target Detection and Tracking

The vision system uses AprilTags as fiducial markers for reliable target identification. The multi-camera setup provides:

  • Wide-angle coverage: 100° field of view per camera
  • Real-time pose calculation: 6-DOF target positioning
  • Coordinate transformation: Camera frame to SLAM coordinate frame
  • Robust tracking: Handles partial occlusion and varying lighting

Motion Planning and Control

Path Planning:

  • Utilizes A* algorithm for optimal path generation
  • Integrates SLAM-generated occupancy grid for obstacle awareness
  • Considers dynamic target movement in planning decisions

Control System:

  • Fine-tuned PID controller for smooth motion execution
  • Differential drive control for precise maneuvering
  • Real-time velocity and heading adjustments

Behavioral State Machine

The FSM manages robot behavior across different operational scenarios:

Primary States:

  1. Search State

    • Systematic environment scanning
    • Target detection within camera field of view
    • Transition to Follow state upon target acquisition

  2. Follow State

    • Continuous target tracking
    • Dynamic path replanning
    • Collision avoidance integration

  3. Recovery State

    • Activated when target leaves field of view
    • Predictive turning based on target’s last known trajectory
    • View expansion maneuvers to reacquire target

State Transitions:

  • Target detected: Search → Follow
  • Target lost: Follow → Recovery
  • Target reacquired: Recovery → Follow

Results and Performance

The system successfully demonstrates autonomous target pursuit with the following characteristics:

  • Reliable detection: Consistent AprilTag recognition across varying conditions
  • Smooth navigation: Collision-free movement in cluttered environments
  • Adaptive behavior: Robust recovery from target loss scenarios
  • Real-time performance: Low-latency response to dynamic target movement

Technical Challenges and Solutions

Challenge 1: Limited Field of View

  • Solution: Multi-camera setup with wide-angle lenses

Challenge 2: Real-time Path Planning

  • Solution: Efficient A* implementation with SLAM integration

Challenge 3: Target Loss Recovery

  • Solution: Predictive FSM with intelligent search patterns

Future Enhancements

Potential improvements for the system include:

  • Advanced Prediction: Machine learning for target trajectory prediction
  • Multi-target Tracking: Simultaneous pursuit of multiple targets
  • Enhanced Sensors: Integration of additional sensor modalities
  • Collaborative Robotics: Multi-robot coordination capabilities

Demo

This project demonstrates the integration of computer vision, autonomous navigation, and intelligent control systems in a practical robotic application.