Dispatching multiple robots in a urban environment

Abstract

Performing verification and validation on robots is necessary to ensure they behave appropriately and safely. However, conducting such tests is labor-intensive and costly. Recent studies show that simulation-based testing for the verification and validation of robotic systems may be a less time-consuming, more economical, and more scalable alternative to testing in the field. Despite the advantages, field testing remains the dominant approach due to 3 main reasons: reality gaps, complexity, and lack of capability. This report aims to address the concerns by developing a realistic simulator by combining Gazebo, a 3D dynamic simulator that is capable of simulating robots in an outdoor complex environment, ROS, an open-source robotics framework able to support the reuse of code, allowing the code used in the virtual robot to be the same as the one used in a real robot, and open street maps, which extracts real-world geographical data to construct the virtual world. With the use of a Docker image, users can run the simulation independently without any dependency or compilation problems. The simulator has a RESTful API, that allows the user to interact with the simulator through the use of HTTP requests. This results in a realistic, simple to use, and with all the capabilities the user requires to accurately and effectively simulate the dispatching algorithm.