In this paper, we present a novel online global reactive motion planner that synergizes the benefits of reactive control and model predictive control (MPC). By applying circular fields, the planner significantly simplifies the problem of determining control inputs for mobile robots and manipulators, making real-time MPC feasible even in complex and dynamic three-dimensional environments. This approach utilizes the performance advantages of optimal control while maintaining reactivity to environmental changes and computational efficiency. The proposed motion planner is evaluated in various simulated scenarios, including complex dynamic environments with up to 100 moving obstacles, and is compared to different state-of-the-art approaches.

For creating robots that are capable of human-like performance in terms of speed, energetic properties, and robustness, intrinsic compliance is a promising design element. In this paper, we investigate the principle effects of elastic energy storage and release for basketball dribbling in terms of open-loop cycle stability. We base the analysis, which is performed for the 1-degree-of-freedom (DoF) case, on error propagation, peak power performance during hand contact, and robustness with respect to varying hand stiffness. As the ball can only be controlled during contact, an intrinsically elastic hand extends the contact time and improves the energetic characteristics of the process. To back up our basic insights, we extend the 1-DoF controller to 6-DoFs and show how passive compliance can be exploited for a 6-DoF cyclic ball dribbling task with a 7-DoF articulated Cartesian impedance controlled robot. As a human is able to dribble blindly, we decided to focus on the case of contact force sensing only, i.e., no visual information is necessary in our approach. We show via simulation and experiment that it is possible to achieve a stable dynamic cycle based on the 1-DoF analysis for the primary vertical axis together with control strategies for the secondary translations and rotations of the task. The scheme allows also the continuous tracking of a desired dribbling height and horizontal position. The approach is also used to hypothesize about human dribbling and is validated with captured data.