RoboCon: A Modular Robotic Containerization, Orchestration, and Load Balancing Technique for Mixed Hierarchy Task Handling Across Computing Platforms

Abstract

This paper introduces an innovative edge-cloud-based modular architecture to enhance the capabilities of multi-robot systems operating in resource-constrained environments. This architecture addresses the limitations often encountered in resource-constrained platforms with lightweight microprocessors by enabling efficient offloading of computationally intensive tasks to hybrid edge-cloud clusters, thereby augmenting computational prowess. Through a comprehensive experimental evaluation, we substantiate the effectiveness of our approach by offloading a rover's perception system to a hybrid edge-cloud cluster, enabling efficient object detection. Our findings highlight the architecture's adaptability, showcasing cloud clusters' speed advantages and local edge clusters' robustness in handling higher network throughput. This insight informs the recommendation to deploy high-throughput applications on edge clusters and computationally intensive tasks with storage and reduced network demands on cloud clusters. Furthermore, this work emphasizes the benefits of containerization and load balancing in modular robotics, facilitating platform-independent development, resource optimization, and enhanced software distribution security. Containerization's minimal impact on CPU and RAM usage has gained prominence in mobile robotic applications. However, supporting mixed-criticality software jobs remains a challenge. This paper presents performance results in federated and adaptive edge-fog-cloud networks, providing valuable insights for planning tasks with mixed-criticality criteria.