Ute Feldmann

Robotic systems are crucial in modern manufacturing. Complex assembly tasks require the collaboration of multiple robots. Their orchestration is challenging due to tight tolerances and precision requirements. In this work, we set up two Franka Panda robots performing a peg-in-hole insertion task of 1 mm clearance. We structure the control system hierarchically, planning the robots’ feedback-based trajectories with a central policy trained with reinforcement learning. These trajectories are executed by a low-level impedance controller on each robot. To enhance training convergence, we use reverse curriculum learning, novel for such a two-armed control task, iteratively structured with a minimum requirements and fine-tuning phase. We incorporate domain randomization, varying initial joint configurations of the task for generalization of the applicability. After training, we test the system in a simulation, discovering the impact of curriculum parameters on the emerging process time and its variance. Finally, we transfer the trained model to the real-world, resulting in a small decrease in task duration. Comparing our approach to classical path planning and control shows a decrease in process time, but higher robustness towards calibration errors.

Robotic systems are crucial in modern manufacturing. Complex assembly tasks require the collaboration of multiple robots. Their orchestration is challenging due to tight tolerances and precision requirements. In this work we set up two Franka Panda robots performing a peg-in-hole insertion task. We structure the control system hierarchically, planning the robots’s trajectories feedback-based with a central policy trained with reinforcement learning. These trajectories are executed by a low-level impedance controller on each robot. To enhance training convergence, we use reverse curriculum learning incorporating domain randomization, varying initial configurations of the task. After training, we test the system in a simulation, studying the impact of curriculum parameters on emerging process characteristics like process time and variance. Finally, we transfer the trained model to a real-world setup, comparing results with simulation as well as classical path planning and control approaches.

We present some parts of our Maths for Engineers teaching concept: above all, motivation: sometimes mathematics teachers observe that engineering students, for example, perceive the maths concepts as a useless evil for them and therefore learn with little motivation and sustainability. When the maths concepts are later used in other subjects, the students do not know how to apply them and teachers do not value the maths education very highly. We solve the motivation problem by linking the mathematics with the subject education: the maths exercise program contains tasks from the subject education at that moment when the necessary maths concepts are available; examples can be found here: https://tud.link/y33m. In this contribution we will document our experience in achieving this maths‐subject‐link and the success by motivating students.Other parts of our teaching concept are: (1.) making recordings of lecture and tutorial available so that students can decide whether to attend the face‐to‐face lecture or watch the video ‘flipped‐classroom‐style’ at home: 4/5 of our students choose face‐to‐face but use the recordings for review. (2.) Instead of TWO small‐group‐tutorials per week, we offer ONE central‐tutorial and ONE small‐group‐tutorial per week: We are documenting the success of this change based on student views and on changes in exam scores: the pass rate did not change, but we are observing a raised midfield. (3.) by means of our extra credit rules: https://tud.link/6r84 we evoke (a) group work, (b) that students continuously ‘stay on task’ throughout the semester, as well as (c) keep reviewing what they have learned so far.