Autonomous Navigation of Rovers Using ML and DL Techniques

Introduction

The applications, and various use-cases of Machine Learning (ML) (Kholiya et al., 2021), Deep Learning (DL) (Pal et al., 2022), and Computer Vision (CV) has led to a revolution in the industrial and scientific applications. Today Artificial Intelligence (AI) (Verma et al., 2022) and ML techniques, and frameworks are used for a wide range of high priority tasks such as scientific computation, autonomous navigation, mechanical and civil engineering requirements, mining, and much more (Alzubi et al., 2018). This boom in the AI/ML frameworks has prompted researchers to use it for space exploration satellites and rovers (Mortlock & Kassas, 2021) to increase the efficiency, and the credibility of the exploration findings for further enhancements under human supervision. Also, quality of data (Bhushan et al., 2017; Bhushan et al., 2018a; Bhushan et al., 2018b; Bhushan & Goel, 2016; Negi & Kaur, 2017) is critical for obtaining better results using AI/ML techniques. It will increase the efficiency, and the credibility of the exploration findings for further enhancements under human supervision. It has expanded the unmanned space exploration missions which are conducted on a large scale. Also, it is not limited to geo-sensing satellites, solar and planetary orbiters (Mortlock & Kassas, 2021), unmanned space vehicles such as lunar and mars rovers, asteroid sample collection bots (Jiang et al., 2020), etc. which aids in the investigation of the immediate solar system. The existing autonomous systems are utilized by aerospace giants for accomplishing various mission parameters, eliminating or minimizing the need for human intervention while assuring the safety parameters.

This chapter describes the use cases of AI/ML in the navigation of rovers and autonomous vehicles. The rovers are used for a variety of purposes; such as planetary exploration, mineral scouting, sample collection, autonomous and remote testing etc. They can be categorized into several types such as mobile robots, agriculture rovers, planetary rovers, unmanned self-driving vehicles, and asteroid sample collectors. The use of a conglomerate of AI/ML frameworks that cover several elements of image recognition, forecasts, strategic planning, and reinforcement of current models instigates this field of research. The DL (Rana & Bhushan, 2022) and Neural Networks (NN) used in the existing works can be tailored and are scalable to suit the custom requirements of the mission or directive given to the rovers. This helps in making optimal measures for any tasks according to the constraints laid out by the limitation of resources which might be different for each and every rover (Alzubi et al., 2018; Daftry et al., 2022). However, it is important to note that while the standard frameworks with their existing parameters may not be much helpful for a more complex task such as one requiring more elaborative processing etc.

Subsequently, it provides various aspects and facilitates provisions for modification using a combination of techniques, and mathematical modelling such as Gaussian Process (Takebayashi et al., 2021), Markov Decision Process (MDP) (Abcouwer et al., 2021)and so on to provide abridgement of data. ML techniques like Linear Regression (LR) (Arrouch et al., 2022)and Multiple Linear Regression (MLR) models can be used to predict limited computation involving problems like determination of the position of solar panels, etc. However, NNs play a significant role as the tasks become increasingly complex involving multiple iterative decisions based on several dynamic variables like object classification and traversal methods.