**1. Introduction**

Over the last few decades, we have observed a rapid development of mobile robotics. Robots such as autonomous vacuum cleaners and lawnmowers are used as standard equipment. The market for autonomous cars and transport within industrial plants is developing.

In 2020, DARPA launched a program called Racer (Robotic Autonomy in Complex Environments with Resiliency) (https://www.darpa.mil/news-events/2020-10-07 accessed on 15 October 2021).

The RACER program aims to develop universal solutions designed to work with various UGV platforms in challenging terrain, taking into account the terrain conditions, at least in terms of adjusting the UGV speed to the environmental conditions. The program aims to develop an autonomy system that will not limit the UGV platform while driving off-road. In other words, if the UGV platform has been designed to drive in rough terrain up to 70 km/h, the autonomy system is to be able to safely navigate the vehicle up to this speed. The program's primary objective is to develop an algorithm that adjusts the local path of the UGV in real time to maximalize the speed of the UGV while driving. Figure 1 presents an example of local path planning with traversability estimation. The algorithm adjusts the path of the UGV to avoid the snowdrift on the front of the platform.

Research on this type of autonomy system at the Łukasiewicz Research Network— Industrial Research Institute for Automation and Measurements PIAP has been conducted since the launch of the project in 2018 called ATENA—Autonomous system for terrain UGV platforms with the following leader function, implemented in the field of scientific research and development work for the defense and security of the state financed by the National Center for Research and Development in Poland under the program Future technologies for defense—a competition of young scientists. The off-road autonomy system was developed and adapted to work with an off-road car. As part of the project, two technology demonstrators were created, functional in following the leader based on

1

the operation of the vision systems. Łukasiewicz—PIAP ATENA system demonstrator has been tested under real conditions in various weather conditions and is fully operative in the following leader function in rough terrain with obstacle avoidance up to 30 km/h. The main advantage of this system is the fact that the vehicle that is following the leader creates the traversability estimation in real time and calculates its own path to follow the leader, in other words, the following vehicle, not just repeating the leader path; it calculates its own adjustment to the set off-road ability. The developed system is equipped with a module for autonomous navigation in an unknown environment. We paid special attention to solving the issue of map construction and planning a collision-free path.

**Figure 1.** Visualization of path planning by a UGV in rough terrain in winter conditions. On the right corner is the image from the roof camera on the UGV.

Figure 2 presents the Łukasiewicz—PIAP ATENA system demonstrators on which the research was performed. The demonstration of the ATENA system contains the off-road car with the Łukasiewicz—PIAP drive-by-wire system and Łukasiewicz—PIAP autonomy controller. The drive-by-wire system allows control of the steering wheel, adjusts the velocity of the vehicle, and controls the brakes, without losing the ability to control the vehicle by a human.

**Figure 2.** Two Łukasiewicz—PIAP ATENA system demonstrators on the test terrain in winter conditions.

This article, which is an extended version of our conference paper [1], presents a method of path planning and building a map of the environment using hexagonal grids. We applied our approach to 3D data obtained using a set of sensors mounted on the ATENA system demonstrator. The data were collected on the premises of the Łukasiewicz Research Network—Industrial Research Institute for Automation and Measurements PIAP. We demonstrate that our algorithm allows us to build accurate models of the environment. The area of the collected data is a rectangle with a circumference of approximately 400 m (Figure 3). The created map is useful for path planning especially in unstructured and rough terrain when the ATENA system demonstrator must search for the lost leader in near historical localization of it or the case of a human leader when the traveling path must be absolutely different than the leader path. The vehicle must calculate its own path not based on the leader path, but based on leader localization.

**Figure 3.** The view of the terrain used in this article.

This paper is organized as follows: After the Introduction and the section discussing related work in Section 3, we briefly describe the mapping module. In Section 4, we present a collision-free path planning algorithm. Section 5 contains experimental results illustrating the advantages of our approach. The article concludes with a summary and bibliography.
