SLAM-based Safe Indoor Exploration Strategy

Omar Mostafa, Nikolaos Evangeliou, Anthony Tzes

TL;DR

This paper introduces a novel strategy for mobile robots to safely explore unknown indoor environments. It combines robust SLAM (Simultaneous Localization and Mapping) with a skeleton-based path planning approach that prioritizes obstacle avoidance while maximizing exploration coverage. This method is particularly effective for non-holonomic robots with limited maneuverability.

What Problem Does This Paper Address?

Exploring unknown indoor environments is a fundamental challenge for mobile robots. Traditional exploration strategies often struggle with cluttered spaces, noisy sensor data, and robots that cannot easily change their orientation or position (non-holonomic constraints). This can lead to inefficient exploration, getting stuck, or even collisions. Existing methods for path planning in complex environments can be computationally expensive, and ensuring both safety and thorough exploration simultaneously is difficult. This paper aims to address this by developing a strategy that is safe, efficient, and robust to real-world sensor noise and robot limitations.

Methods

The core of the proposed strategy lies in a two-pronged approach:

  • Robust SLAM: The system utilizes RTAB-Map, an open-source SLAM library, which integrates data from a 3D LiDAR and an RGB-D camera. IMU (Inertial Measurement Unit) data is also incorporated to improve localization accuracy and reduce drift, especially in environments with repetitive features. RTAB-Map builds a 2D occupancy grid map of the environment.
  • Skeleton-Based Safe Exploration: Instead of traditional frontier-based exploration, this method leverages a "skeleton" representation of the free space. This skeleton represents the core pathways a robot can navigate. The exploration algorithm then focuses on identifying "safe headings" that lead to unexplored areas while maximizing the distance from detected obstacles. This is achieved by analyzing the robot's forward-facing LiDAR scans and computing allowable regions for movement. The algorithm prioritizes headings that are furthest from obstacles and lead to new areas.
  • Waypoint Selection and Navigation: Once a safe heading is determined, the robot moves to a new waypoint at a fixed distance along that heading. The system uses the Navigation2 software, enhanced with the skeleton-based path planning, to navigate to these waypoints. The approach also includes mechanisms to handle situations where forward motion is blocked, allowing for reverse maneuvers and trying alternative safe angles to avoid getting stuck.

Datasets

The paper does not explicitly mention training on specific datasets. Instead, it focuses on experimental validation using a real-world mobile robot in an obstacle-rich apartment environment. The data used for evaluation is generated by the robot's onboard sensors (LiDAR, RGB-D camera, IMU) during these experiments.

Key Insights and Ideas

  • Skeleton as a Safety Net: Using a skeleton representation of the environment, derived from the occupancy grid, provides a more robust and computationally efficient way to identify navigable paths compared to dense graph representations.
  • Prioritizing Safety and Exploration: The strategy elegantly balances the need for safety (staying away from obstacles) with the goal of exploration (finding new areas). The "safe heading" computation is key to this balance.
  • Robustness to Non-holonomic Robots: The method is specifically designed for robots with non-holonomic constraints (like the dual-wheeled robot used), which are common in indoor robotics but pose challenges for traditional path planners.
  • Leveraging Existing SLAM: By building upon a mature SLAM system like RTAB-Map, the exploration strategy benefits from accurate mapping and localization, which are crucial for effective exploration.

Key Results

The paper presents experimental validation with a self-balanced, dual-wheeled robot in a real-world apartment. The results demonstrate:

  • Successful Exploration: The robot was able to explore a cluttered apartment, navigate through multiple rooms, and successfully exit enclosed spaces.
  • Robustness to Noise: The system effectively handled noisy sensor measurements and avoided positional drifts, maintaining an accurate map.
  • Safe Navigation: The robot consistently prioritized obstacle avoidance while expanding its explored territory.
  • Qualitative Demonstration: Figures in the paper visually showcase the robot's path planning and the generated map, illustrating the algorithm's effectiveness in a complex environment.

Limitations and Future Work

Limitations:

  • The paper focuses on 2D exploration.
  • The distance to the next waypoint is kept constant, which might not be optimal in all scenarios.
  • The "safe heading" computation relies on the accuracy of the LiDAR's forward-facing scans.

Future Work:

  • Extending the exploration framework to multi-robot systems for collaborative exploration and unified map construction.
  • Further optimizing the path planning to dynamically adjust waypoint distances.
  • Investigating 3D exploration capabilities.

Summary updated

BibTeX
@inproceedings{mostafa2025slam,
  author = {Omar Mostafa and Nikolaos Evangeliou and Anthony Tzes},
  title = {{SLAM-based Safe Indoor Exploration Strategy}},
  booktitle = {ICARA},
  year = {2025},
  doi = {10.1109/icara64554.2025.10977630},
}
Figure