TAIS - WP3

SAFE AND SECURE AUTONOMOUS ROBOTS FOR SENSITIVE ENVIRONMENTS

The development of humanoid robotics for the introduction into sensitive environments such as medical or pediatric environments may reveal sensitive information that should not be processed outside that environment. This work package focuses on the early childhood education context and the possibility of employing AI applications that interact with pre-elementary school pupils. Recently, movements have been worldwide to ensure trustworthiness, privacy, and fairness in autonomous systems (those capable of taking actions with little or no human supervision) and, generally, any technology that uses automatic decision-making via artificial intelligence algorithms (He, 2021; Hesken, 2021).

These movements have impacted the assistive robotics industry; nowadays, there is no straightforward solution to the issue of trustworthiness from the top market players (Chatzimichali, 2021). For instance, Softbank Robotics and its famous humanoid robot Pepper have a closed environment for their applications, making it difficult to audit them to ensure their fairness and lack of biases. Other famous humanoid robots such as Amy and Sanbot are permanently connected to servers in mainland China, making the situation on privacy and trust to the use of robotic technologies even worse (Hägglund, 2022). These situations are usually the rule and not the exception within the industry. This conundrum could impact technologies' future development and use in potentially sensitive environments (childcare, hospitals, etc.). It can also be considered a problem by different market actors. The issue of trustworthiness also opens the door to develop and study new technologies that could lead to a new generation of autonomous robots. At Arcada, we have experience in designing these services on humanoid robots, as we have access to three humanoid robots and other sets of non-humanoid ones (Majd, 2021). We have developed know-how in the last few years via different projects that allow us to include non-proprietary applications within their operative systems (Hägglund, 2022).