Improving Human-Robot Object Exchange by Online Force Classification

Robots are sometimes tasked with handing an object over to a human, which can be a challenging task for a robot to perform, especially when the human partner has no experience in interacting with robots. This paper presents our work to enable a robot to learn how to achieve this task with wrist force/torque sensing. Firstly, we present a device to record the data, then we discuss techniques used for teaching. We focused on the classification problem as defined in our paper to enable the robot to find the finger-opening movement. The main challenge is that the classification should be run online at a comparable rate to the controller. To achieve a computationally efficient classifier, we used the Wavelet Packet Transformation for feature extraction, and then we used the Fisher criterion to reduce the dimension of features. A Relevance Vector Machine is used for continuous classification procedure mainly for its sparsity. Some recorded data and results from dimension reduction are shown. We then discuss the accuracy and sparsity of classification by the Relevance Vector Machine in this application. The software of continuous classification on forces is then tested on the robot for interactive object exchange between human and robot, which gives promising results.