The HeAL9000 project aims to develop a “service robot” for the healthcare sector. The aim is to develop a platform that provides cognitive capabilities in a robot capable of: (i) learning information through perception of the environment in which it operates, (ii) interacting with people and objects in the environment in order to care for patients in a natural way, (iii) monitoring contextual events and actions to be triggered in the event of specific contingencies.
The platform
The software platform supporting the deployment of this robotic agent will be responsible for managing the lifecycle (from training to validation) of the Machine Learning and Process Mining processes that enable the cognitive functionalities outlined above, in order to improve them based on information obtained in the field and adapt them to the context of use and the user.
The pre-industrial prototype will design a complex ‘cognitive agent’, capable of anticipating the user’s needs based on physiological/environmental monitoring and providing assistance both in terms of responding to adverse situations and in terms of physical/cognitive stimulation to improve the quality of therapy. The proposed architecture must ensure the installation and updating of new validated models via OTA (Over The Air) in order to support the robot’s intelligence in an evolutionary manner and align its characteristics with research and studies in the sector.
The robot, with its sensors and computational capabilities, must adapt its operating model to the ecosystem in which it is embedded. It will be capable of learning the habitual behaviours of the individual with whom it interacts, the actions that the individual must necessarily perform, and monitoring the main parameters detectable directly or through interaction via environmental sensors, wearable sensors and specialised devices. The robot will therefore be able to analyse the volume of information collected in the field (sensing) (i) directly on site or (ii) using an external service (e.g. in the cloud); at the same time, it will be able to carry out direct actions, either by deciding on necessary actions (act) or by determining the appropriate responses to the questions of the monitored subject: these actions may involve autonomous actions or responses/actions directed at other connected devices, thereby acting as an intelligent gateway to them.
Objectives
The objectives pursued within the project are therefore:
- To study the mechanisms of therapist-patient interaction in order to make robot-mediated rehabilitation treatment more effective than the current state of the art. We will analyse the mechanisms underlying multimodal interaction between therapist and patient during traditional treatment, with the aim of developing technologies capable of replicating them in robot-assisted rehabilitation.
- Develop a system for verbal communication and physical and cognitive interaction. In order to replicate the modes of communication and multimodal interaction typical of the therapist-patient dyad, a patient interface system will be developed that integrates a speech synthesis system for verbal communication with the patient, a vision system for recognising the patient’s gestures and facial expressions, and a system for monitoring the patient’s biomechanical and psychophysiological state.
