"Bionic" platform
Manager: Abderrahmane KHEDDAR
The bionics platform supports research and development at the interface between robotics, in particular humanoid robotics and biomimetics, and musculoskeletal rehabilitation and regeneration.
Passive knee unloading orthesis
Léa Boillereaux thesis (LMGC, LIRMM)
Supervisors: Simon Le Floc'h, Frank Jourdan, Abderrahmane Kheddar
The development of a wearable brace for knee unloading that is as passive as possible, with a focus on unloading intra-articular knee forces in squat movements. The choice fell on the use of pneumatic cylinders and a cam that controls the movement of the piston in the cylinder body to create the desired force at each angle of flexion. The cam profile is designed according to patient data. The initial pressure to be imposed on the cylinders is used to set the percentage of contact force discharge during movement. A high initial pressure (i.e. at 0° flexion-extension) will therefore be imposed on the cylinder immediately after implantation of the neo-tissues, then gradually reduced over the weeks of rehabilitation. We also decided to use robotics to test the effectiveness of our orthosis. Using a Panda Franka Emika robot combined with a 3D-printed setup reproducing a human leg, we reproduce the fine kinematics of the knee while applying forces. Force sensors are placed at strategic points on the assembly (notably under the tibial plateau). The brace is then mounted on the setup, and the force applied by the robot can be easily compared with the force measured at the tibial plateau. The difference absorbed by the orthosis is a measure of its effectiveness.
Active prosthesis control from physiological EMG signals
Louise Scherrer thesis (LIRMM, CHU - IRMB)
Supervisors: Abderrahmane Kheddar, Christian Jorgenessen
The goal of this project is to develop an exoskeleton control system based on sEMG (surface) sensors, and to make a passive exoskeleton active by means of signal processing and control. The first part of the project consists in developing learning or classification methods to identify the instants and phases when the movement of a patient (the user of the prosthesis) requires additional torque assistance. This part identifies the control inputs for an active prosthesis. Validation of the control and AI part is achieved by motorizing a passive exoskeleton and testing the transparency of the prosthesis/patient interaction.
Development of a 3D pressure sensor for interaction measurement
Postdoc Youcan Yan (LIRMM)
Supervision : Abderrahmane Kheddar
Implantation of artificial cartilage (direct bioprinting) is a highly promising treatment in the research phase. However, current trials run the risk of failing to keep the artificial cartilage alive after implantation, as existing knee braces lack tactile sensing capabilities comparable to those of skin, and therefore cannot progressively release the supporting forces applied to the knee during the rehabilitation phase of the artificial cartilage. To this end, we are developing a tactile 3D pressure sensor technology that can measure force distributions (in normal and tangential directions) on the knee tibia (emulated by a two-armed robot), and can also be used for interaction forces between a knee brace and human skin. This technology can help to design, evaluate and control the knee braces of an artificial leg (emulated by a two-armed robot) so that the brace is comfortable to wear and the support forces generated by the brace can be actively adjusted in line with the progress of artificial cartilage rehabilitation.
Real-time imaging of knee prostheses
Collaboration BoneTag - LIRMM
Collaborators: Sylvain Dutrieux, Arnaud Tanguy, Stéphane Naudi, Abderrahmane Kheddar
In this project, two robots (Franka Emika) are used in coupling, one holding a 3D print of the tibial plateau fitted with part of a metal knee prosthesis, and the other holding a 3D print of part of the femur on which the other part of the prosthesis is mounted. The robots are then programmed to generate all theoretically possible movements of the two parts of the prosthesis, and in parallel collect data from the BoneTag sensors and the direct kinematics of the two robots. The robots provide an exact relative position of the two prostheses in the human knee, and thus a real-time image of the implanted prosthesis. The robot data is used to learn the imaging rendering by an artificial intelligence that learns to interpret the BoneTag sensor signals from the robot kinematics. Current work is focused on trying to image the knees and breasts using small temporary metal implants.
