Hypergrid: A Hyper-Fast ROS-Based Framework for Local Map Generation

One of the main tasks of an autonomous vehicle is being able to navigate from one point to another, while constantly perceiving and considering every element in the environment. Path planning techniques are divided into global and local planning, being the latter the ones responsible for tracking the dynamic obstacles presented around the vehicle. These methods often rely on the occupancy grid map (OGM) data structure to represent the environment due to its low complexity, which the Robotic Operating System (ROS) provides through the costmap_2d package in the navigation stack. Although this package is widely used in the robotics community, it struggles to keep up with the information provided by a multisensor setup from an autonomous vehicle, which provides a large amount of three-dimensional data. For this reason, this paper presents Hypergrid, a ROS-based framework with GPU and CPU acceleration to process raw sensor data and generate local OGMs useful for navigation tasks. Furthermore, real-world experiments on different autonomous driving platforms are performed, in order to validate the proposed approach and compare it with the performance of costmap_2d. The source code of Hypergrid is available in github.com/lsi-uc3m/hypergrid under GPLv3 license.