Artificial Intelligence for satellite management: the HMI challenge

Artificial Intelligence for satellite management: the HMI challenge

According to the United Nations Office for Outer Space Affairs, there are over 8,000 satellites in orbit, and just over half of them are active. This traffic leads to several implications in the sky – for instance, increasing the risk of collision between satellites, as the result of the formation of space debris – and in satellite control stations, where operators manage an increasing amount of data in highly complex situations. As is already the case in Air Traffic Management (ATM), Artificial Intelligence can support satellite management – including the ability to monitor multiple satellites simultaneously – saving time and resources, and ensuring effectiveness and safety. As with any application that requires interaction and teaming between humans and AI, the success of the partnership will depend on the ability to design intuitive and more usable human-machine interfaces (HMI) for users.


Artificial Intelligence in space: an overview of applications

If robots ever join astronauts on space missions, hopefully they won’t be as depressed as Marvin, the paranoid android from Douglas Adams’ novel “The Hitchhiker’s Guide to the Galaxy,” a robot so human that he has a strong personality, despite being one of the worst. Is it early to think about sending robots into space? “Let’s say that it’s started to be talked about, especially out of necessity: for instance, delivering new robot assistants that would be similar to cobot models already implemented in manufacturing operations: that would reduce mission time and costs and increase astronaut safety” explains Simona Turco, Head of Business Development at Deep Blue. “Space agencies and industry are working on it, to ensure that research leads to operational experimentation in the coming years”.

Cobots aside, astronauts already use robotic systems (i.e. remote-controlled vehicles, manipulator arms, mobility aids) in extravehicular activities, such as the assembling or the maintenance of structures, and manoeuvrability of hazardous materials; but we have to state that they are human-controlled robotic systems, not automated ones. Despite the belief that autonomous systems haven’t made their way into space operations, the reality is quite the opposite: even Artificial Intelligence, the most advanced form of automation, has been implemented in space technology. “Space probes for planetary exploration already use AI to make autonomous decisions, for example, to manage the unforeseen event of trajectory changes, or to search for specific rock samples, and to sift through the huge amount of data they collect in order to select only useful information to onshore monitoring centres.”

The most interesting – and versatile – applications of Artificial Intelligence in space concern satellites. “In recent years, AI has been successfully implemented to improve various aspects of satellite operation: such as communication, data processing, prediction of the percentage of collision risk with space debris or other satellites, – continues Turco – on the other hand, despite its potential, AI has not been fully exploited concerning the support for monitoring and control of satellites from Earth stations, in order to assist the operator in making prompt and corrective interventions based on priority scales”. 


Artificial Intelligence in terrestrial satellite control centres

Nowadays, satellite management is remotely performed from consoles located in terrestrial control centres, where the operator can simultaneously monitor and intervene on each individual satellite, or a group set in true multi-satellite systems. In the case of a multi-satellite system, the operator manually selects data from the subsystems settings of each satellite in the constellation, and switches between different interfaces associated. “The priorities of interventions must be evaluated by the operator considering the potential consequences of not taking corrective action. Therefore, regardless of the level of autonomy of the individual satellites, it is clear that in multi-satellite systems the monitoring of activities requires managing an increasing amount of information, in complex and diversified operational scenarios”, says Turco.

In the wake of the multiple and digital remote towers implemented in air traffic management systems, Artificial Intelligence could assist in the management of multiple satellites. “The AI would introduce a new preliminary data analysis work, ‘suggesting’ the satellite on which the operator should intervene and the type of intervention to be implemented according to real-time data analysis and the specific needs. Thus, it would help both to reduce complexity of the information to be managed and to prioritize data to be processed, finally improving the decision-making process”.


Artificial Intelligence needs valuable human-machine interfaces

In addition to algorithms for satellite data analysis and prioritization, the integration of Artificial Intelligence into satellite management procedures means delivering intuitive and “easy-to-read” interfaces, despite the complexity of datas’ representations (a task that must involve Human Factors, as we have discussed here and here). “The main goal of intelligent interfaces is to improve user-machine interactions, hence the operator’s performance”, explains Stefano Bonelli, Head of Innovative Human Factors in Deep Blue.

In this field, the research is particularly focused on the so-called adaptive interfaces, which are designed to adapt to the user’s cognitive structure (i.e. their way of thinking) and their mental and emotional state. “The design of adaptive interfaces is easy for all those applications with few data and scenarios to manage. But if the complexity increases, as in the case of multi-satellite management, the amount of information available increases as well; thus, the interfaces must be capable of presenting all this data to users in an understandable way, in order to avoid stressful workload and attention drops”.

On paper, adaptive interfaces are the ultimate and innovative applications of intelligent automation, they have an intrinsic problem though: by autonomously deciding which information to deliver and when – based on a change in the user’s state or external conditions – they modify the system’s behaviour and partially take away the operator’s ability to predict how the system will behave, leading to a potential loss of trust in the system and paradoxically increasing effort and stress. They can also represent several issues when they are managed by multiple operators simultaneously, due to the naturally different cognitive structures of each individual. “The design and implementation of adaptive automation represent a great challenge for research – says Bonelli – but these innovative interfaces have the great advantage of constantly keeping users in contact with essential information, integrating the machine’s ability to record actions and automate tasks with the understanding of context and the logical actions of humans”. For these reasons, this is a challenge that we cannot afford to lose.


Get in touch with us