Distributed ledger technology such as blockchain is considered essential for supporting large numbers of micro-transactions in the Machine Economy, which is envisioned to involve billions of connected heterogeneous and decentralized cyber-physical systems. This stresses the need for performance and scalability of distributed ledger technologies. Addressing this, sharding techniques that divide the blockchain network into multiple committees are a common approach to improve scalability. However, with current sharding approaches, costly cross-shard verification is needed to prevent double-spending. This paper proposes a novel and more scalable distributed ledger method named ScaleGraph that implements dynamic sharding by using routing and logical proximity concepts from distributed hash tables. ScaleGraph addresses cybersecurity in terms of integrity and availability to support frequent micro-transactions between autonomous devices. Benefits of ScaleGraph include a total storage space complexity of O(𝑡), where t is the global number of transactions (assuming a constant replication degree). This space is sharded over N nodes so that each node needs O(𝑡/N) storage in expectation, which provides a high level of concurrency and data localization as compared to other delegated consensus proposals. ScaleGraph allows for a dynamic grouping of validators that are selected based on a distance metric. We analyze the consensus requirements in such a dynamic setting and show that a synchronous consensus protocol allows shards to be smaller than an asynchronous one, and likely yields better performance. Moreover, we provide an experimental analysis of security aspects regarding the required size of the consensus groups with ScaleGraph. Our analysis shows that dynamic sharding based on proximity concepts brings attractive scalability properties in general, especially when the fraction of corrupt nodes is small.

In truck manufacturing, an assembly line is typically used to produce many models and variations of trucks in any desired order. Thus, stations are fed with specific sets of parts, known as kits, depending on what truck is next in line. This paper focuses on how to automate the layout planning for placing parts on a kitting wagon. Layout planning resembles the pallet loading problem, but differences include that there is no layering, there may be constraints on how each part can be placed (orientation) and there may be predefined layout hints suggesting positions. A layout planner based on constraint programming is presented. The objective is to facilitate a decision support loop where an engineer may add placement hints as constraints (e.g., for enhancing the workflow at assembly stations) and reuse placements from similar kits to provide recognition. The layout planner automatically generates layout proposals. Finally, it is the engineer that approves the layout plan. There may be many acceptable solutions that have different scores (i.e., levels of quality). We show some approaches to reduce the search space to improve performance. The evaluations show the score and time needed for finding results on some problem instances that are optimally solvable. In the general case, however, finding an optimal solution may be unnecessary or even intractable.

The efficiency of an assembly line depends on how the work is distributed along the line. This is known as the Assembly Line Balancing Problem, an NP-hard optimization problem. Automatic solvers for this problem have been studied for decades but have not been widely adopted in the industry, resulting in a theory-practice gap. The typical automation approach assumes that all constraints and objectives are known and can be statically defined ahead of time such that solvers with a precisely defined objective function can take a fully specified problem instance as input and produce a (near) optimal solution as output. In some industries, meeting these assumptions is particularly challenging because of properties such as mixed-model production with high model variance, multi-manned stations, large task graphs, etc. This paper explains why, in certain industries, such as automotive end assembly, complete automation is likely infeasible in practice due to challenges in modeling the problem, collecting data, and specifying the objective function. Manual intervention by an engineer as a decision-maker is therefore unavoidable. We argue that maximizing automation, by helping the decision-maker be as effective as possible, requires a decision support system (DSS) that supports an interactive and iterative workflow, thereby enabling assisted planning. Furthermore, we identify solver features that become relevant in the DSS context, thus making the case that focusing on standalone solvers, and treating the integration into a DSS as an implementation detail, is not a viable option. We conclude that decision support systems play a central role in closing the theory-practice gap.

Many optimization problems cannot be solved to optimality in practice due to time constraints. Instead, the goal is to find sufficiently good solutions in reasonable time. This creates a trade-off between running time and solution quality. Some problems require iterative and interactive solving due to uncertainties in the definition of the problem instance or objective function. These present an additional challenge because the desired trade-off can vary between iterations. We focus on simple assembly line balancing, an NP-hard combinatorial optimization problem, as a use case. Absolute termination criteria, such as a time limit or iteration/generation limit, are often inadequate due to their rigidity. We therefore evaluate the effect of more flexible termination criteria (e.g. unimproved time limit and score limit) on the performance of a solver, focusing on the ability to reduce unproductive time and dynamically achieve an adjustable time-quality tradeoff. Our results show that it is possible to simultaneously reduce running time and increase solution quality on average. While quality improvements are modest, time savings can be substantial. In some cases, running time can be reduced by as much as 55 % without sacrificing solution quality.

A mixed-model assembly line produces a variety of product models on a single assembly line. Since models differ e.g. in work content, some stations may be unable to finish their tasks in time if multiple work-intensive models are placed consecutively on the assembly line, resulting in costly overload situations. This leads to the product sequencing problem, an NP-hard optimization problem that consists in arranging orders into a suitable sequence. While sequences are typically decided and frozen some time before production to enable logistics planning, late changes are sometimes unavoidable, which has received little attention. This paper presents the early stages of developing a decision support system for handling such late changes.