UAV-as-a-Service for Robotic Edge System Resilience

摘要

By melding the capabilities of robotics with the agility of edge computing, Robotic Edge System (RES) exemplifies the next generation of Internet Intelligent Service Systems, delivering incredible efficiency and adaptability across diverse real world applications. Inevitably, RES is susceptible to mechanical disruptions of robots, particularly when some tasks are assigned to faulty ones, leading to uncertain failures and performance degradation. Due to communication and latency constraints, it is not always feasible to rely on edge/cloud computing infrastructure for system recovery. To address these issues, a UAV-as-a-Service (UAVaaS) approach is proposed that leverages the mobility of UAVs to enhance system resilience. Specifically, a Markov Decision Process (MDP) is utilized to assign tasks dynamically among active UAVs to achieve system recovery in a livestock monitoring scenario. Additionally, a Dual Noise Deep Deterministic Policy Gradient (DNDDPG)-based mechanism is proposed to minimize system recovery time and energy consumption. The proposed DNDDPG enhances exploration and decision-making during training by integrating parameter noise and behavioral noise into the classic Deep Deterministic Policy Gradient (DDPG) algorithm. The simulation results indicate that the proposed mechanism can achieve convergence within 100 episodes, thereby effectively minimizing the time and energy required for system recovery.