Blockchain Brings Trust to Collaborative Drones and LEO Satellites: An Intelligent Decentralized Learning in the Space

Abstract

In this paper, we develop a foundation for a constellation of Low Earth Orbit (LEO) satellite IoT by constructing a Blockchain-based framework for continual knowledge sharing and learning collaboratively. This approach is directly applicable for a swarm of Unmanned Aerial Vehicles (UAVs). We ablate Federated Learning (FL) successful features as a basis to ensure high precision of learning inferences at timescales relevant to the underlying time-varying space network and channel dynamics. In such a dynamic setting, there is always a likelihood that miners may be compromised or fail to propagate information in time because of some intrinsic factors such as channel impairments, satellite handovers and attacks. Such transmission failures often lead to undesirable forking events in the Blockchain. Consequently, maintaining a low energy consumption and smallish delay in such an erratic network is highly nontrivial and challenging. To quantify the impacts of the forking and minimize the occurrence of such unwanted events and their adverse effects, we develop a procedure to estimate the expected energy consumption for a given set of miners, block transmissions, and LEOs’ or UAVs’ mobility. Besides, we shed light on deep learning-based resource allocation for mobile mining and demonstrate the synergic gain of FL with Blockchain.