• Applied machine learning for soft object contact

As the brain of modern smart robots, the machine learning algorithm tremendously expands the robotic applications in the past few years. Supported by NSERC Alliance, we are developing a deep learning algorithm for the haptic robotic teleoperation interacting with soft objects. Via training, our algorithm is able to estimate the complex interaction force with soft
objects which is extremely challenging to measure accurately. Based on a robust estimation of the interaction forces with soft objects, the robot will in further detect the hardness of the objects, and make “smart” feedback for intelligent haptic teleoperations. Our approach will greatly ease high-level human-machine interactions. We have made significant achievements in estimating interaction forces with different soft objects by touching using low training data.