Artificial intelligence in the space industry

The beginning of AI and how it has grown over time

In the early 1990s, networks of artificial neurons, the first trace of AI, appeared in some development processes for Ariane launchers. Inspired by the way the human brain works, these particularly powerful mathematical models could rapidly perform repetitive, time-consuming and complex tasks.

In 2018, ArianeGroup put in place an AI-dedicated area of study. It was divided into four branches: machine learning uses available data to best harness the company’s extensive experience in space launchers; knowledge management, made popular by language models like ChatGPT, makes it easy to collect, organizes, and distribute data in order to improve organisational performance; multi-agent systems are based on interactions between several AIs; and operational research advances new strategies, which may sometimes appear counterintuitive, to solve clearly identified problems. The purpose of all four is to provide myriad applications that can optimally help engineers in their day-to-day tasks.

A bespoke assistant for engineers
To develop specialized tools and be sure to comply with ArianeGroup’s high reliability and quality standards, engineers use open-source algorithms to build suitable solutions for every area of business. Each implementation is approved by technical experts who determine if the algorithms meet security norms for space launchers, and, if applicable, adhere to fundamental laws of physics. In meeting these requirements, engineers provide themselves with an effective data control assistant that has a complete view of all data, and which can retrieve useful information faster than the human brain.

AI has led to real progress: processing and analysis of large quantities of data, enhanced precision, continuous learning, and automation. Nevertheless it has its limits. AI remains an assistant with no decision-making capabilities. It simply analyses huge quantities of data, and reports anomalies. Engineers make the actual decisions; their expertise is vital. By providing them with assistance in their day-to-day tasks, AI means they can spend more time on high value-added creative activities for on-going and new projects.

Application examples
Today, AI plays a role in every stage of a product’s life cycle. For example, it can improve robustness by detecting anomalies in parts on production lines. Quickly replacing a defective part can save valuable time. Saving time in the design processes of a launcher is also an advantage. AI-run simulators can rapidly generate predictive behavioural models, through aerodynamic calculations or shock propagation within the vehicle. Another important aspect not to be overlooked is documentation. Language models have greatly accelerated and simplified its processing, analysis, and production.

The new dynamic around AI means cultural change, which, in turn, requires training to best understand, take ownership of, and perfect existing tools. ArianeGroup engineers must be able to determine where AI can help, and where they should maintain control. It will take a few more years for launchers to have embedded AI systems that can detect problems and make decisions. The idea is currently being investigated by the ENLIGHTEN (European iNitiative for Low cost, Innovative & Green High Thrust ENgine) project. Led by ArianeGroup, and funded by the European Commission, the ENLIGHTEN aims to accelerate development of key technologies such as AI for the operational monitoring and maintenance of future reusable launcher engines. Stay tuned!