Autonomous planetary vehicle development platform

Abstract

This paper presents the design, architecture, and constructions of a planetary autonomous exploration vehicle platform, which can be used to develop and test Artificial Intelligence, based software and generate a large dataset for the training of neural networks. Rovers will be at the frontier of planetary exploration, capable of executing tasks without human supervision in a harsh and unpredictable environment, and to do so it requires real-time command execution to keep it away from a risky situation. Due to the limitations imposed by communication latency and small window to communicate through deep space satellites, existing mars rovers are semi-autonomous. To develop AI-based software for the rover, a low-cost alternative of a planetary rover is required to generate data from different types of sensors and actuators for a long duration of time and perform all possible scenarios and actions. Presently this task is done using simulation or replicas of the actual rovers used in planetary missions which are very costly. The proposed rover design is a low-cost alternative, capable of powering, driving varieties of sensors, scale up to new hardware and record data as specified by the user. It can also be used to test the newly developed algorithm before being tested on an actual rover. This platform can be used as a simulation platform for software as the proposed platform is directly in contact with the environmental factors.