Robot Mapping, Estimation, and Interaction

Autonomous navigation system with SLAM capabilities for TurtleBot2

Project Overview

Solo development of a comprehensive autonomous navigation software stack for TurtleBot2 platform, focusing on real-time sensor fusion and mapping capabilities. This project demonstrates advanced robotics algorithms including SLAM, path planning, and sensor integration.

Duration: January 2024 – May 2024
Role: Solo Developer
Platform: TurtleBot2 with Ubuntu Linux
Institution: University of Rochester Supervisor: Thomas M. Howard

Technical Implementation

System Architecture

  • Platform: TurtleBot2 robot with differential drive
  • Operating System: Ubuntu Linux
  • Framework: ROS (Robot Operating System)
  • Programming Language: C++
  • Sensors: LiDAR, IMU, wheel encoders

Key Algorithms Implemented

SLAM (Simultaneous Localization and Mapping)

  • Algorithm: Extended Kalman Filter (EKF) SLAM
  • Performance: 10% reduction in localization error compared to baseline
  • Features: Real-time pose estimation and map building

Mapping

  • Method: Occupancy Grid Mapping
  • Resolution: Configurable grid resolution for different environments
  • Updates: Probabilistic map updates using sensor measurements

Path Planning

  • Algorithm: A* search algorithm
  • Performance: 15% reduction in planning time
  • Features: Dynamic replanning, obstacle avoidance

Key Achievements

  • Developed and deployed a complete C++ software stack for autonomous navigation
  • Implemented real-time sensor fusion combining LiDAR, IMU, and odometry data
  • Achieved robust navigation in uncertain and dynamic environments

Performance Improvements

  • Localization: 10% improvement in pose estimation accuracy
  • Planning: 15% reduction in path planning computation time
  • Reliability: 95% success rate in reaching navigation goals

Software Engineering

  • Designed comprehensive unit-testing framework for sensor integration
  • Implemented fault detection and recovery mechanisms
  • Created modular, reusable software components

Real-time Performance

  • Met strict real-time requirements for mobile robot navigation
  • Optimized algorithms for embedded computing constraints
  • Maintained 20Hz control loop frequency

Results & Validation

Experimental Setup

  • Environment: Indoor laboratory with static and dynamic obstacles
  • Test Scenarios: Multiple navigation tasks with varying complexity
  • Metrics: Localization accuracy, planning time, success rate

Performance Metrics

  • Localization Error: < 5cm RMS in structured environments
  • Planning Time: < 100ms for typical scenarios
  • Navigation Success: 95% goal reaching success rate
  • Real-time Performance: Consistent 20Hz operation

Future Enhancements

  • Multi-robot Coordination: Extend to collaborative SLAM
  • Dynamic Obstacles: Improved tracking and prediction
  • Semantic Mapping: Integration of object recognition
  • Outdoor Navigation: GPS integration for larger environments