Simon Manschitz, Berk Güler, Wei Ma, Dirk Ruiken, "Sampling-Based Grasp and Collision Prediction for Assisted Teleoperation", International Conference on Robotics and Automation (ICRA), 2025.

Shared autonomy allows for combining the global planning capabilities of a human operator with the strengths of a robot such as repeatability and accurate control. In a real-time teleoperation setting, one possibility for shared control is to let the human operator decide for the rough movement and to let the robot do fine adjustments, e.g., when the view of the operator is occluded. We present a learning-based concept for shared autonomy that aims at supporting the human operator in a real-time teleoperation setting. At every step, our system tracks the target pose set by the human operator as accurately as possible while at the same time satisfying a set of constraints. An important characteristic is that the constraints can be dynamically activated and deactivated which allows the system to provide task-specific support. Since the system generates robot commands in real-time, solving an optimization problem in every iteration is not feasible. Instead, we sample target configurations and use Neural Networks for predicting the constraint costs for each configuration. By evaluating each configuration in parallel, our system is able to greedily select the target configuration which (approximately) satisfies the constraints and has the minimum distance to the operator's target pose. We evaluate the framework in simulation and demonstrate that it can be utilized for teleoperating a Franka Emika 7 degrees of freedom Panda robot with a Robotiq gripper.