[Paper Review] A-star path planning simulation for UAS Traffic Management (UTM) application

Carlos Augusto Pötter Neto, Gustavo de Carvalho Bertoli+1 more|arXiv (Cornell University)|Jul 27, 2021

Robotic Path Planning Algorithms35 references4 citations

TL;DR

This paper presents a ROS/Gazebo-based simulation platform for UAS Traffic Management (UTM) that implements 2D and 3D A* path planning with Manhattan and Euclidean heuristics. It demonstrates that using Manhattan distance reduces execution time by up to 17.2× compared to Euclidean distance and that map reduction decreases average execution time by 84.5×, significantly improving performance for urban UTM applications.

ABSTRACT

This paper presents a Robot Operating System and Gazebo application to calculate and simulate an optimal route for a drone in an urban environment by developing new ROS packages and executing them along with open-source tools. Firstly, the current regulations about UAS are presented to guide the building of the simulated environment, and multiple path planning algorithms are reviewed to guide the search method selection. After selecting the A-star algorithm, both the 2D and 3D versions of them were implemented in this paper, with both Manhattan and Euclidean distances heuristics. The performance of these algorithms was evaluated considering the distance to be covered by the drone and the execution time of the route planning method, aiming to support algorithm's choice based on the environment in which it will be applied. The algorithm execution time was 3.2 and 17.2 higher when using the Euclidean distance for the 2D and 3D A-star algorithm, respectively. Along with the performance analysis of the algorithm, this paper is also the first step for building a complete UAS Traffic Management (UTM) system simulation using ROS and Gazebo.

Motivation & Objective

• To develop a scalable, simulation-based platform for UAS Traffic Management (UTM) using open-source robotics tools.
• To evaluate the performance of 2D and 3D A* path planning algorithms under different heuristic functions (Manhattan vs. Euclidean).
• To analyze the impact of map representation resolution on A* execution time and route optimality.
• To establish a foundation for a complete, extensible UTM simulation system supporting multi-drone operations and regulatory compliance.
• To support future integration of dynamic constraints, scheduling, and advanced path planning algorithms.

Proposed method

• The simulated urban environment was generated from an occupancy map and decomposed into a grid-based node representation for path planning.
• Two versions of the A* algorithm were implemented: one for 2D paths at discrete altitudes and another for full 3D route computation.
• The Manhattan and Euclidean distances were used as heuristic functions to estimate cost-to-goal in the A* search process.
• The path planning system was integrated into a ROS-based framework using custom ROS packages for navigation and simulation.
• Map reduction was applied by grouping 5×5 nodes into single super-nodes to reduce computational load.
• Performance was evaluated by measuring route length and execution time across multiple start-goal pairs on both original and reduced maps.

Experimental results

Research questions

• RQ1 How does the choice of heuristic (Manhattan vs. Euclidean) affect the execution time and route quality of A* in 2D and 3D path planning for UAS in urban environments?
• RQ2 What is the performance impact of map resolution on A* path planning execution time and route optimality?
• RQ3 How does the use of a reduced map representation affect the scalability of A* for large urban UTM simulations?
• RQ4 Can the A* algorithm be effectively integrated into a ROS/Gazebo-based UTM simulation platform for real-time drone navigation?
• RQ5 What are the trade-offs between route optimality and computational efficiency in A* path planning for UTM applications?

Key findings

• The 2D A* algorithm with Manhattan heuristic executed 3.2× faster than with Euclidean heuristic, while the 3D version showed a 17.2× performance gain.
• Using a reduced map representation decreased average A* execution time by 84.5× compared to the original 200×200 grid, despite a 7.5% increase in route length.
• The route length was on average 8% shorter when using the original map, but the performance gain from map reduction far outweighed this minor loss in optimality.
• Statistical analysis (Wilcoxon signed-rank test) confirmed significant differences in execution time between heuristic functions and map representations, with p-values < 0.001.
• The 3D A* algorithm with Manhattan heuristic showed the best balance of speed and route quality, making it most suitable for real-time UTM simulation.
• The simulation platform successfully demonstrated the feasibility of integrating A* path planning into a ROS/Gazebo-based UTM framework, forming a foundation for multi-drone and dynamic constraint extensions.