A blockchain-based LLM-driven energy-efficient scheduling system towards distributed multi-agent manufacturing scenario of new energy vehicles within the circular economy

As processing technology is becoming increasingly complex, a single enterprise is no longer able to satisfy all the customization needs, which also requires extra energy consumption and vast time to seek cooperation from other enterprises for processing, especially in the field of new energy vehicle manufacturing. To coincide with the circular economy principle, a blockchain-based Large Language Model (LLM)-driven energy-efficient scheduling solution is proposed towards distributed multi-agent manufacturing scenario of new energy vehicles in this work. Firstly, distributed manufacturing system has the potential to efficiently organize distributed manufacturing resources by abstracting various machines from different factory nodes into agents with corresponding processing capabilities. Additionally, energy-efficient scheduling in line with the circular economy principles helps to optimize production cycle and reduce energy consumption and delay time, thereby lowering production costs and enhancing competitiveness. Compared with the traditional methods that suffer from long training time and local optimization, LLMs offer innovative solutions by learning a wealth of experiential knowledge in advance from vast amounts of data to further support self-adaptive and real-time energy-efficient scheduling. It is also worth noting that untrusted production data in factories may mislead the learning process of LLM, which may generate incorrect decision results. Therefore, a credit evaluation-based consensus mechanism is proposed to provide a trustworthy data access in distributed manufacturing, which can improve the transparency and traceability of the whole production process. Finally, the proposed approach is validated in the distributed manufacturing scenario for new energy vehicles. Compared with common methods, experimental results demonstrate the superiority of the proposed method on the distributed multi-agent manufacturing scenario for new energy vehicles, highlighting its potential to enhance production efficiency and circular economy.