Leo Explore

Leo Explore

Project Objective

The objective of this project was to design and implement an autonomous reconnaissance robot capable of exploring unknown environments using a modular and intelligent navigation system. The robot is designed to operate in both simulated and real-world environments, with applications ranging from disaster response to military reconnaissance.

Contributors

Abstract

This project introduces the Reconnaissance Bot, a robotic system built to navigate and explore uncharted environments with minimal human intervention. The bot uses a top-down control architecture composed of three key components:

The Frontier Server identifies potential exploration goals using a K-Means clustering algorithm to process frontier points. The Global Planner employs the A* algorithm to compute an optimal path from the robot’s current location to the selected goal. The Local Planner implements a Model Predictive Path Integral (MPPI) control method to generate smooth, collision-free trajectories within the robot’s immediate environment.

The system integrates seamlessly with ROS-based SLAM for localization and mapping, allowing a clean separation of responsibilities between mapping and planning modules. This decoupled architecture ensures modularity, extensibility, and robust real-time performance. The Reconnaissance Bot was evaluated through extensive simulation using Gazebo and real-world testing on the Northeastern University campus.

Demo Video

Results & Analysis

Simulation Testing

Two Gazebo simulation environments were used to evaluate system performance:

TurtleBot3 House — A hexagon-shaped grid map resembling a turtle.
TurtleBot3 World — A more realistic, house-like indoor setup.

The primary metric used to assess performance was the Rate of Exploration, defined as the ratio of explored area to time taken (m²/s). The simulation revealed that the exploration rate decreases non-linearly as the environment size increases, following an inverse proportionality to the 1.5th power of the area.

Environment	Area Explored (m²)	Time Taken (s)	Rate of Exploration (m²/s)
TurtleBot3 House	21.88	74s	0.30
TurtleBot3 World	83.82	560s	0.15

Real-World Testing

The system was deployed in two real-world environments at Northeastern University:

4x4 Maze World
Hurtig Hall Corridor

The local planner, operating at 12 Hz, successfully generated agile, real-time trajectories using MPPI. Perturbation-based trajectory sampling enabled smooth obstacle avoidance, while velocity control (0.8 m/s nominal, angular range [0, π] rad/s) facilitated natural movement. Demonstration videos captured during testing validated the planner’s effectiveness in real-world conditions, closely matching the simulation results.

Overall, the Reconnaissance Bot consistently performed seamless exploration and robust navigation across a variety of structured environments, with modular code and scalable planning strategies providing a strong foundation for future enhancements.