Self-Supervised Path Planning in UAV-aided Wireless Networks based on Active Inference

Secondly, we use the learned This paper presents a novel self-supervised path-planning method world model as an internal generative model enriched with active for UAV-aided networks. First, we employed an optimizer to solve states to simulate the environment and plan actions that minimize training examples offline and then used the resulting solutions as the agent's surprise during online decision-making. This approach demonstrations from which the UAV can learn the world model to enables the UAV to navigate its surroundings with a reference model understand the environment and implicitly discover the optimizer's representing the goal, choosing actions that minimize unexpected or policy. UAV equipped with the world model can make real-time unusual observations (surprise) measured by how much they deviate autonomous decisions and engage in online planning using active from the expected goal. The main contributions of this paper are as inference. During planning, UAV can score different policies based follows: It expands on previous research [11] by exploring online on the expected surprise, allowing it to choose among alternative planning, a prospective form of cognition.