In this position paper, the authors argue the need for a novel framework that provides flexibility, autonomy and optimizes the use of scarce resources to ensure reliable communication during next-generation space missions. To this end, the authors present the shortcomings of existing space architectures and the challenges in realizing adaptive autonomous space-networking. In this regard, the authors aim to jointly exploit the immense capabilities of deep reinforcement learning (DRL) and cross-layer optimization by proposing an artificial intelligence-based cognitive cross-layer decision engine to bolster next-generation space missions. The presented software-defined cognitive cross-layer decision engine is designed for the resource-constrained Internet-of-Space-Things. The framework is designed to be flexible to accommodate varying (with time and location) requirements of multiple space missions such as reliability, throughput, delay, energy-efficiency among others. In this work, the authors present the formulation of the cross-layer optimization for multiple mission objectives that forms the basis of the presented framework. The cross-layer optimization problem is then modeled as a Markov Decision Process to be solved using deep reinforcement learning (DRL). Subsequently, the authors elucidate the DRL model and concisely explain the deep neural network architecture to perform the DRL. This position paper concludes by providing the different phases of the evaluation plan for the proposed cognitive framework.
Artificial Intelligence-based Cognitive Cross-layer Decision Engine for Next-Generation Space Mission
2019-06-01
281621 byte
Conference paper
Electronic Resource
English
ARISE - The next generation space VLBI mission
AIAA | 1995
|SimNGST: The Next Generation Space Telescope Mission Simulator
British Library Conference Proceedings | 1998
|