This thesis discusses two subjects: distributed ledgers scalable enough to support micro-transactions, and automating assembly planning in manufacturing industries.

Established blockchain solutions achieve high robustness and reliability. By being distributed and decentralized, they avoid a single point of failure, and fault-tolerant consensus mechanisms ensure that the system works as intended in the presence of malicious participants. However, their main weakness is scalability. The two most popular and well-known blockchain solutions require all nodes to store all transactions, and the transaction throughput is far too low to compete with traditional, centralized transaction processing systems. To improve scalability, systems have been developed that split the network nodes into groups that can process transactions in parallel, also known as sharding. We propose a novel approach to sharding, called ScaleGraph, which uses concepts from distributed hash tables to define shards dynamically. Nodes store and validate only transactions involving those accounts that are close to the node according to a logical distance metric. This greatly reduces the storage burden on each node and allows any number of transactions involving distinct accounts to be validated in parallel. By employing a synchronous consensus protocol, shards can be kept smaller without sacrificing fault tolerance, especially in a permissioned environment where participation can be controlled and a small fraction of malicious nodes can be assumed.

Manufacturing is a highly complex process in many industries, involving many different planning problems, and increasing automation has the potential to make manufacturing more efficient. This thesis presents a proof-of-concept solution to the kitting layout planning problem, where a list of parts has to be placed on a kitting wagon for delivery to an assembly line station. However, some problems have proven difficult to automate in practice, despite decades of research. One such problem, assembly line balancing, is analyzed in depth. We identify fundamental challenges that make the goal of complete automation implausible in some industries, such as automotive manufacturing. Human intervention is thus unavoidable, suggesting that assisted planning is a more promising approach for achieving increased automation in practice. Therefore, we make the case that bridging the gap between theory and practice requires decision-support systems that enable an iterative and interactive workflow.

This thesis addresses two research areas: scalable distributed ledgers for micro-transactions, and the automation of assembly planning in manufacturing industries.

Established blockchain solutions are robust and reliable. Being distributed and decentralized, they avoid a single point of failure, and fault-tolerant consensus mechanisms ensure that the system works as intended even when some participants are faulty or malicious. However, their main weakness is scalability. The two most popular and well-known blockchain solutions, Bitcoin and Ethereum, require all nodes to store all transactions, and their transaction throughput is far too low to compete with traditional, centralized transaction processing systems. To improve scalability, systems have been developed that split the network nodes into groups that can process transactions in parallel, a technique known as sharding. We propose a sharded system called ScaleGraph that uses a novel architecture with one transaction per block and one shard per account, designed to maximize parallelism. The design is inspired by concepts from distributed hash tables, particularly to define shards based on a logical distance metric for node IDs and account IDs. Nodes store and process only transactions involving those accounts that are close to the node according to the distance metric. This greatly reduces the storage burden on each node and allows any number of transactions involving distinct accounts to be validated in parallel. We also design a new cross-shard transaction commit protocol for this architecture. The protocol offers global serializability and inevitable atomic commit, without the need for an abort path. This is achieved using only shard-local consensus and certificate exchange, rather than global or joint cross-shard consensus.

Manufacturing is a highly complex process in many industries and involves many different planning problems where increasing automation has the potential to make manufacturing more efficient. This thesis presents a proof-of-concept solution to the kitting layout problem, where a list of parts has to be placed on a kitting wagon for delivery to an assembly line station. However, some problems have proven difficult to automate in practice, despite decades of research. One such problem, assembly line balancing, is analyzed in depth. We identify fundamental challenges that make the goal of complete automation implausible in some industries, such as automotive manufacturing. Human intervention is thus unavoidable, suggesting that bridging the gap between theory and practice requires decision support systems for assisted, iterative, and interactive planning. The thesis also includes preliminary work on the product sequencing problem, limited to framing the use case, assumptions, and requirements. Subsequent ongoing work suggests strong parallels to assembly line balancing, indicating that the identified challenges and possibilities for addressing them reflect a broader pattern in industrial planning automation.