top of page
4094CD61-7D04-4D2A-BFC3-55D4B20D1866.png

Robot Navigation with Localization & Path Planning  

I developed an autonomous robotic navigation system designed to operate in complex, maze-like environments with unknown or partially known layouts. The system focused on adaptive path planning, obstacle avoidance, and real-time localization, using a combination of probabilistic filtering and geometric planning techniques. The robot was tasked with navigating toward a goal while detecting and adjusting to the presence of wooden obstacles, simulating indoor rescue or exploration scenarios.

 

Implemented in Python using NumPy, SciPy, and Matplotlib, the system fused particle-filter localization with A* path planning to determine optimal trajectories while maintaining an accurate belief of the robot’s position. Ultrasonic sensors were used for distance measurement and obstacle detection, providing real-time updates to the internal map and localization model. 

Localization: Detecting Current Positon

  • Trained the robot to determine its current position using particle filtering (probabilistic filtering)

​

  • This ensures that the robot knows its starting point before calculating the path to the final position

​​

  • Trained the robot on a circular track with cardboard box markers to learn its relative position before deploying it in the maze

A* Path Planning & Maze Navigation

  • Modeled the workspace (physical environment with obstacles) of the robot.

  • Derived the configuration space (C-space) to account for robot dimensions and collision boundaries.

  • Designed and visualized obstacle inflation to ensure paths respected clearance margins.

  • Applied the L2 (Euclidean) distance heuristic to guide the search efficiently while guaranteeing optimality.

  • Implemented the A* path planning algorithm to compute the shortest path from the starting position to the goal

Navigating The Maze

Robot Capabilities:

 

  • Uses ultrasonic sensors to detect obstacles 

  • Determines its position by sensing surrounding obstacles through localization with particle filtering

  • Apply the A* path planning algorithm to compute the shortest collision-free path from start position to the goal

  • Navigates to the goal using the planned path and odometry 

​

*Odometry: The process of estimating a robot’s position by measuring its own movements (e.g., wheel rotations or motion data)

bottom of page