Control Barrier Functions and Lidar-Inertial Odometry for Safe Drone Navigation in GNSS-denied Environments

Halil Utku Unlu, Dimitris Chaikalis, Vinicius Gonçalves, Anthony TzesIntechOpen 2023 · DOI · OpenAlex

Abstract

This chapter is concerned with drone navigation in unknown, indoor environments. This necessitates using the onboard LiDAR and IMU sensors to solve the simultaneous localization and mapping (SLAM) problem. Control barrier functions (CBFs) augmented with circulation constraints are designed for motion planning. CBFs ensure that the drone can safely navigate the unknown environment by avoiding obstacle collisions. The FAST-LIO package is used for SLAM and the generated OctoMap data are transmitted to the CBF-module motion planning algorithm. Simulation studies using the Gazebo Physics Engine with a coaxial hexarotor drone are provided to validate the efficacy of the suggested algorithm.

Introduction

Navigating a drone indoors presents a significant challenge in robotics, often described as a "triple threat": the absence of GNSS signals, the requirement for real-time mapping of unknown environments, and the necessity of guaranteed safety during goal-oriented movement. Traditional path planners, such as RRT* or PRM, typically rely on pre-existing maps, which are unavailable during initial exploration. Furthermore, reactive obstacle avoidance methods frequently encounter "local minima," where the drone may stall or oscillate when its path is obstructed. This research aims to bridge the gap between high-level exploration and low-level safety-guaranteed control.

System Architecture and Methodology

The proposed navigation framework is structured into three primary modules: SLAM, Global Planning, and Local Planning.

Figure 2: System architecture block diagram showing the interaction between SLAM, Global Planner, and Lower-Level Controller.
  1. SLAM Module: The system utilizes FAST-LIO, a computationally efficient LiDAR-Inertial Odometry package. It fuses 3D LiDAR points with IMU data through a tightly coupled Iterated Extended Kalman Filter (IEKF). This data generates an OctoMap (a 3D occupancy grid). For planning efficiency, a specific height-slice of this 3D map is projected into a 2D grid.
  2. Global Planner: This module manages exploration by identifying "frontier points"—the boundaries between known free space and unknown territory. An incremental Exploration Graph tracks connectivity, and a cost metric determines whether the drone should proceed toward the global goal or explore a frontier to bypass obstacles.
  3. Local Planner: The core innovation lies in the use of Control Barrier Functions (CBFs) formulated as Quadratic Programs (QP). To prevent stagnation at local minima, the CBFs are augmented with circulation constraints. By employing a skew-symmetric matrix, the planner creates a "vector field" that encourages the drone to circulate around obstacles rather than simply stopping.

Key Insights

A pivotal aspect of this approach is the use of circulation to solve stagnation. While standard CBFs act as a safety "shield," they can lead to deadlocks if a goal is directly behind a wall. The addition of a tangent circulation vector mathematically encourages the drone to "slide" along obstacle boundaries until an opening is found.

Figure 3: Visualization of the 3D LiDAR point cloud (left) and its corresponding 2D occupancy grid projection (right).

Additionally, the use of a conservative 2D projection from the 3D OctoMap balances computational speed with safety for indoor missions. A graph-based memory system further prevents the "forgetfulness" common in purely reactive systems, allowing the drone to backtrack from dead ends to promising frontier points.

Experimental Validation

The framework was validated using the Gazebo Physics Engine in a complex warehouse simulation. The simulated platform was a coaxial hexarotor equipped with a 3D LiDAR, IMU, and downward-facing rangefinder.

Results demonstrated successful navigation from start to goal across a complex environment. The system autonomously switched between goingToGlobal and goingToFrontier modes as needed. Crucially, the CBF-based controller maintained a safety radius of 0.9m from all obstacles, even with real-time map updates at 100Hz.

Conclusion and Future Work

While effective, the current 2D constraint assumes a constant flight altitude. Future work will explore full 3D CBF implementations for more complex environments, such as multi-story buildings. Additionally, extending the framework to handle dynamic obstacles and transitioning from simulation to physical hardware—addressing real-world sensor challenges like glass reflections—are key next steps.

Summary updated

Abstract

This chapter is concerned with drone navigation in unknown, indoor environments. This necessitates using the onboard LiDAR and IMU sensors to solve the simultaneous localization and mapping (SLAM) problem. Control barrier functions (CBFs) augmented with circulation constraints are designed for motion planning. CBFs ensure that the drone can safely navigate the unknown environment by avoiding obstacle collisions. The FAST-LIO package is used for SLAM and the generated OctoMap data are transmitted to the CBF-module motion planning algorithm. Simulation studies using the Gazebo Physics Engine with a coaxial hexarotor drone are provided to validate the efficacy of the suggested algorithm.
BibTeX
@inproceedings{unlu2023control,
  author = {Halil Utku Unlu and Dimitris Chaikalis and Vinicius Gonçalves and Anthony Tzes},
  title = {{Control Barrier Functions and Lidar-Inertial Odometry for Safe Drone Navigation in GNSS-denied Environments}},
  booktitle = {IntechOpen},
  year = {2023},
  doi = {10.5772/intechopen.1002654},
}
Figure