Purpose: I worked on developing an autonomous mobile robot using ROS to navigate a maze environment using SLAM, AMCL localization, and LiDAR data. A major challenge arose when similar-looking sections of the maze caused localization ambiguity, requiring extensive debugging in RViZ to analyze pose estimation errors and particle filter behavior. By the end of the course, we implemented improvements that significantly increased localization reliability and overall navigation performance, though some edge cases remained imperfect. 

Objectives:

1. Implement SLAM to generate an accurate map of an unknown maze environment

2. Use AMCL for probabilistic localization based on LiDAR sensor data

3. Develop autonomous navigation using ROS navigation stack

4. Diagnose and resolve localization errors caused by environmental symmetry

5. Validate system performance through repeated maze navigation trials

Mapping & Localization

• SLAM: Implemented using ROS with LiDAR data to generate a 2D occupancy grid map of an unknown maze environment.

• AMCL Localization: Used a particle filter–based approach to estimate robot pose within the generated map; tuned sensor and motion model parameters to improve convergence.

• RViZ Debugging: Visualized particle clouds, costmaps, and TF frames in RViZ to diagnose pose drift and ambiguity caused by symmetric maze features. 

Navigation

• Path Planning: Utilized the ROS navigation stack with global and local planners to compute collision-free paths to target goals.

• Obstacle Avoidance: Integrated LiDAR-based costmaps for dynamic obstacle detection and safe trajectory execution.

• Goal Execution: Deployed autonomous waypoint navigation with continuous localization feedback. 

Localization Robustness Improvements

• Symmetry Handling: Identified failure cases where similar maze regions caused multi-modal particle distributions and incorrect pose estimates.

• Parameter Tuning: Adjusted AMCL noise parameters and scan matching thresholds to reduce divergence.

• Validation: Repeated navigation trials to evaluate pose accuracy and recovery behavior after localization errors. 

System Integration

• ROS Architecture: Structured nodes for mapping, localization, planning, and control with proper topic communication and TF frame management.

• Testing Pipeline: Established a workflow of map generation → localization → goal setting → navigation → performance evaluation. 

Lab Report