Search

Systems-of-systems (SoS) often include multiple agents that interact in both cooperative and competitive modes. Moreover, they involve multiple resources, including energy, information, and bandwidth. If these resources are limited, agents need to decide how to share resources cooperatively to reach the system-level goal, while performing the tasks assigned to them autonomously. This paper takes a step toward addressing these challenges by proposing a dynamic two-tier learning framework, based on deep reinforcement learning that enables dynamic resource allocation while acknowledging the autonomy of systems constituents. The two-tier learning framework that decouples the learning process of the SoS constituents from that of the resource manager ensures that the autonomy and learning of the SoS constituents are not compromised as a result of interventions executed by the resource manager. We apply the proposed two-tier learning framework on a customized OpenAI Gym environment and compare the results of the proposed framework to baseline methods of resource allocation to show the superior performance of the two-tier learning scheme across a different set of SoS key parameters. We then use the results of this experiment and apply our heuristic inference method to interpret the decisions of the resource manager for a range of environment and agent parameters. [ABSTRACT FROM AUTHOR]

The vision of Industry 4.0 is to materialize the notion of a lot-size of one through enhanced adaptability and resilience of manufacturing and logistics operations to dynamic changes or deviations on the shop floor. This article is motivated by the lack of formal methods for efficient transfer of knowledge across different yet interrelated tasks, with special reference to collaborative robotic operations such as material handling, machine tending, assembly, and inspection. We propose a meta reinforcement learning framework to enhance the adaptability of collaborative robots to new tasks through task modularization and efficient transfer of policies from previously learned task modules. Our experiments on the OpenAI Gym Robotics environments Reach, Push, and Pick-and-Place indicate an average 75% reduction in the number of iterations to achieve a 60% success rate as well as a 50%-80% improvement in task completion efficiency, compared to the deep deterministic policy gradient (DDPG) algorithm as a baseline. The significant improvements achieved in the jumpstart and asymptotic performance of the robot create new opportunities for investigating the current limitations of learning robots in industrial settings, associated with sample inefficiency and specialization on one task through modularization and transfer learning. [ABSTRACT FROM AUTHOR]

Modern engineered systems are increasingly connected, with complex interactions among social and technical aspects, both during the design process and after fielding. For example, autonomous vehicles, smart cities, sharing economy platforms, and more broadly, emerging systems are enabled by artificial intelligence (AI) and the Internet of things, cross traditional cyber and physical, machine and human, boundaries with their performance largely determined by how they interact with the society. In the light of these trends, sociotechnical systems can be defined by the dynamic interaction between technical systems (hardware, software, and AI agents) and human agents (at the individual, social, and organizational levels). They are often multi-stakeholder, highly heterogeneous in terms of the needs and behavior of users and stakeholders, and are only weakly controllable, due to the presence of different types of autonomy within the system. These factors result in high levels of complexity, cast new. [ABSTRACT FROM AUTHOR]

Modern engineered systems are increasingly connected with complex interactions among social and technical aspects, both during the design process and after fielding, for example, autonomous vehicles, smart cities, sharing economy platforms, and more broadly, emerging systems enabled by artificial intelligence (AI) and the internet of things, cross traditional cyber and physical, machine and human, boundaries with their performance largely determined by how they interact with society. While traditional engineering design has focused on designing an optimal technical artifact, there is an increasing recognition that social and organizational aspects of how designers collaborate and create, and how systems co-evolve with the human and built environments through use, are equally important drivers of value. Additionally, the rise of new sources of data and increased availability thereof create many opportunities to extend design research into the sociotechnical realm, for example, collecting and learning from new sources of data,... [ABSTRACT FROM AUTHOR]