QoS and Energy-Aware Run-time Adaptation for Mobile Robotic Missions: A Learning Approach
Résumé
Mobile robotic systems are normally confronted with the shortage of on-board resources such as computing capabilities and energy, as well as significantly influenced by the dynamics of surrounding environmental conditions. This
context requires adaptive decisions at run-time that react to the dynamic and uncertain operational circumstances for guaranteeing the performance requirements while respecting the other constraints. In this paper, we propose a reinforcement learning based approach for QoS and energy-aware autonomous robotic mission manager. The mobile robotic mission manager leverages the idea of reinforcement learning by monitoring actively the state of performance and energy consumption of the mission and then selecting the best mapping parameter configuration by evaluating an accumulative reward feedback balancing between QoS and energy. As a case study, we apply this methodology to an autonomous navigation mission. Our simulation results demonstrate the efficiency of the proposed management framework and provide a promising solution for the real mobile robotic systems.
Mots clés
mapping parameter configuration
energy-aware autonomous robotic mission manager
Robot sensing systems
Reinforcement learning
Measurement
run-time adaptation
quality of service
power aware computing
mobile robots
learning (artificial intelligence)
path planning
autonomous mobile robots
control engineering computing
learning approach
energy efficiency
environmental conditions
on-board resources
mobile robotic systems
navigation
reinforcement learning based approach
QoS
energy-aware run-time adaptation
mobile robotic mission manager
energy consumption
autonomous navigation mission
accumulative reward feedback