Dual-Time-Scale Cloud-Edge-End Collaborative Task Offloading for Multi-AGV Intelligent Warehousing in Industrial...

Area of Science:

• Industrial Internet of Things (IIoT)
• Robotics and Automation
• Distributed Computing

Background:

• Multi-AGV intelligent warehousing faces challenges with latency-sensitive tasks due to limited onboard resources.
• Purely local processing is insufficient for real-time needs, while cloud offloading causes significant transmission delays and overhead.
• A cloud-edge-end collaborative architecture is needed to balance local and remote processing capabilities.

Purpose of the Study:

• To investigate the joint optimization of AGV service-point migration and task offloading in IIoT warehousing.
• To minimize long-term accumulated system delay while adhering to task latency and AGV energy constraints.
• To develop an efficient algorithm for coordinating AGV movements and task distribution decisions.

Main Methods:

• A dual-time-scale optimization model was formulated to consider service-point selection and offloading impacts.
• The DPSO-MAPPO algorithm was proposed, integrating a discrete particle swarm optimization (DPSO) for service-point planning and multi-agent proximal policy optimization (MAPPO) for task offloading.
• The algorithm operates on slow and fast time scales, with feedback enabling coordination between movement and offloading decisions.

Main Results:

• The DPSO-MAPPO algorithm demonstrated stable convergence in simulations.
• System delay was reduced by 13.55% compared to benchmark algorithms.
• Improvements were observed in total energy consumption and the control of energy violations.

Conclusions:

• The proposed joint optimization and DPSO-MAPPO algorithm effectively addresses the challenges of real-time task processing in multi-AGV intelligent warehousing.
• The cloud-edge-end collaborative approach significantly enhances system performance by balancing computational load and communication overhead.
• This method offers a viable solution for improving efficiency and energy management in advanced IIoT environments.