Development of adaptive safety constraint by predicting trajectories of closest points between human and co-robot

Safety is a critical component for human–robot cohabitation. The control barrier function (CBF) provides an effective tool to build up the constraint and ensure the safety of human–robot interaction. However, since the human and robot keep moving during human–robot interaction, the closest points between parts of them also change. Especially, the human trajectories are not known in prior, which may cause the above safety constraint to fail. In this paper, we construct the safe constraints based on discrete control barrier function (DCBF) by redefining the distance between each link of the robot and each part of the human body as the distance between two line segments in the space. In addition, the look-backward-and-forward strategy is applied to update the neural network model for predicting of human’s motion trajectory effectively. Meanwhile, the root mean square estimation error is included in the safe constraints as the metric of uncertainty to compensate the estimation error of the predicted trajectory. Combining the discrete-time control Lyapunov function, a comprehensive control method under human–robot-coexistence environment is formed. The trajectory of a human’s right arm collected by Qualisys capture system. The experiment are set up by integrating above testbed with a virtual KUKA iiwa model built by MATLAB. The results show that the robot can maintain a safe distance from the human when the DCBF-based constraints with prediction information are used, which verifies the effectiveness of the proposed method.