During the first period, the ethical guidelines to be adopted within AIDE were defined. After that, the target scenarios were identified by end-users’ focus groups. Specifically, four scenarios emerged gathering the user requirements of the different focus groups: 1) communication; 2) Environmental control; 3) Hygiene task; and 4) Preparing and eating a meal. Based on end-users’ requirements in each scenario, AIDE system were characterized and specified. One of the main results of the first period was the prototype of a wearable and wireless system for biosignal recording and processing to be used as a key element in the AIDE multimodal sensory processing system.
The main outcomes of the second year are:
- Cedar Foundation obtained the ethical approval for testing AIDE system. In addition, three experimental sessions were conducted to monitor the design and developments of AIDE hardware/software components.
- A modular standard architecture for AIDE multi-modal interface based on the messaging system Yet Another Robotic Platform (YARP) has been developed, completed and tested.
- A novel shoulder-elbow robotic exoskeleton attached to a wheel chair fixed system with the possibility of using it for either the left or the right arm, according to the user's residual motion capabilities. To complete AIDE exoskeleton, a new prono-supination and hand assistance exoskeleton has been developed, integrated and tested.
- A universal system to control the movements of electric wheelchairs linked with AIDE interfaces has been developed and tested.
- Two intention detection methods based on: 1) hybrid EEG/EOG system; and 2) EMG activity.
- Algorithms to detect and track the 3D position and orientation of texture-less objects and Gaze estimation algorithms.
- Algorithms to recognize the user activity in real ADL tasks: Support Vector Machine (SVM), Artificial Neural Network (ANN) and Decision Tree (DT) algorithms.
- Development and optimization of the Low Level Controller of AIDE arm exoskeleton.
- A motion planning system grounded on a Learning by Demonstration approach and Dynamic Movement Primitives (DMP) was developed.
-The Finite-State Machine (FSM) was enriched with additional modules, to give the user full control of the task execution.
- A first version of the HLC and LLC for communication, control and entertainment was developed.
During the third period, the main outcomes are:
-An operative prototype of the AIDE exoskeleton was delivered and obtained all the certifications needed for be tested with human subjects.
-The implementation of user’s intention detection realized by fusing the physiological signals of EEG and EOG was successfully accomplished and documented in a published paper in Nature Scientific Reports (Figure 2).
-EMG analysis focused on the onset detection algorithm was developed, implemented in real time and tested with healthy subjects.
-To estimate and compute the current position of the users in their home setting, two solutions have developed and tested: 1) based on WiFi Fingerprinting techniques; and 2) based on laser range-finders.
-The performance of the developed algorithms to estimate the 3D pose of texture-less objects and gaze point can be ranked as optimum regarding KPIs of the project.
-Algorithms to recognize the user activity in real ADL tasks were tested and achieved an accuracy about 87%.
-Hierarchical Learning control architecture was completed and tested. The achieved results showed a 100% success rate in the task fulfillment, with a high level of generalization with respect to the environment variability.
-The finite-state machine (FSM) for the AIDE shared control system was enriched with new features (Figure 2).
-A successful integration of the final prototype.
-The AIDE validation testing was completed onsite at Cedar Foundation premises in Belfast Northern Ireland with a total of 17 participants
