Pathfinding in hierarchical representation of large realistic virtual terrains for simulation systems

Highlights

• Solution for pathfinding in large realistic terrains used in simulation systems.

• Uses a hierarchical pathfinding algorithm and virtual terrain representation.

• Employed in the prototype of a military simulation with massive virtual terrain.

• Compared to other hierarchical and non-hierarchical techniques.

• Speeds up the search and scales up well in massive virtual terrains.

Abstract

A prominent pathfinding challenge is the fast computation of path plans in large virtual terrain environments used in simulation systems. Based on a hierarchical approach for pathfinding and terrain representation, this work details the design, the implementation and the test of a quadtree-based structure in the irregular grid navigation map representation of large realistic virtual terrain environments. To achieve this goal, it is described how a hierarchical global A* pathfinding algorithm searches for a path in a coarse initial irregular grid structure then proceeding with the search in refined regions of interest where obstacles are found. Thus, the proposed solution presents as major contribution the enhanced quadtree-based map representation for pathfinding in large virtual terrains. The pathfinding response time of this solution is statistically compared with different hierarchical/non-hierarchical and regular/irregular terrain representation structures instantiated in the modeling of “small” realistic terrain scenarios. After these experiments, similar ones are developed in a massive (large) virtual terrain inserted into a real-life simulation system for the development of military tactical training exercises. The results show that the response time required to generate pathfinding results can be reduced when the proposed global pathfinding technique over the quadtree-based hierarchical and irregular navigation map representation of the large virtual terrain is explored in the development of simulation systems.

Introduction

The work presented in this paper is motivated on the research and development of a real-world military simulation system (the SIS-ASTROS simulation system) for the systematic development of virtual tactical simulation-based training exercises (SIS-ASTROS, 2014). The SIS-ASTROS is a simulation system to train military personnel in the tactical employment of artillery batteries over a battlefield scenario. There, different characteristics of the scenario are considered when selecting the best locations to deploy such agents, as well as to displace their constituent elements in the best possible doctrine-based organization over the terrain, according to the requirements of a given mission. This simulator considers very large military scenarios, which are over 2500 km², constructed in a highly realistic way via computer graphics techniques and the use of GIS-based map data regarding real-world terrains (e.g., Backes et al., 2017, Engel and Pozzer, 2016, Frasson et al., 2018). A realistic and effective global navigation and obstacle avoidance of the batteries’ elements in these large virtual terrain environments is a fundamental requirement to preserve the fidelity of the resulting simulations, providing the user the desired sense of immersion in the simulation-based training exercises developed. To fulfill such requirements, the simulation system's response time to compute path routes for simulated agents is a key issue, and has to be less than 1 s.

To address the presented time-framed pathfinding issue, the SIS-ASTROS simulation system relies on a hybrid semi-autonomous navigation framework (Brondani et al., 2017, Brondani et al., 2018) in which the proposed hierarchical global pathfinding solution is integrated with force-based local navigation techniques (Reynolds, 1999). While local navigation algorithms are detailed elsewhere (e.g., Menezes & Pozzer, 2018), this work is focused on the design, implementation and test of the quadtree-based navigation map structure and the hierarchical pathfinding algorithm components of the SIS-ASTROS simulation system, aiming at providing reliable support to the development of time-efficient pathfinding calculations. In this research field driven by the development of real-world AI applications, the key contribution of this paper is the proposal of the quadtree-based hierarchical and irregular structure as the navigation map representation for large realistic virtual terrains and, based on this contribution, this paper describes the hierarchical adjustment of the A* pathfinding algorithm to perform time-efficient path computations in such massively large virtual terrains as an additional minor contribution.

The paper is organized as follows: Section 2 reviews terrain topologies. Section 3 describes the proposed hierarchical and irregular virtual terrain representation. Section 4 the adapted global A* algorithm to work in the proposed virtual terrain representation. Section 5 describes the performed experiments and discusses the results. Finally, Section 6 concludes the work and presents directions for future work.