当前位置: X-MOL 学术Auton. Robot. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Autonomous assembly planning of demonstrated skills with reinforcement learning in simulation
Autonomous Robots ( IF 3.7 ) Pub Date : 2021-10-16 , DOI: 10.1007/s10514-021-10020-x
Joris De Winter 1, 2, 3 , Ilias EIMakrini 2, 3 , Greet Van de Perre 2, 4 , Ann Nowé 2, 5 , Tom Verstraten 2, 3 , Bram Vanderborght 2, 4
Affiliation  

Industrial robots used to assemble customized products in small batches require a lot of reprogramming. With this work we aim to reduce the programming complexity by autonomously finding the fastest assembly plans without any collisions with the environment. First, a digital twin of the robot uses a gym in simulation to learn which assembly skills (programmed by demonstration) are physically possible (i.e. no collisions with the environment). Only from this reduced solution space will the physical twin look for the fastest assembly plans. Experiments show that the system indeed converges to the fastest assembly plans. Moreover, pre-training in simulation drastically reduces the number of interactions before convergence compared to directly learning on the physical robot. This two-step procedure allows for the robot to autonomously find correct and fast assembly sequences, without any additional human input or mismanufactured products.



中文翻译:

在仿真中使用强化学习对展示技能进行自主装配规划

用于小批量组装定制产品的工业机器人需要大量重新编程。通过这项工作,我们的目标是通过自主寻找最快的装配计划来降低编程复杂性,而不会与环境发生任何冲突。首先,机器人的数字孪生在模拟中使用健身房来了解哪些组装技能(通过演示编程)在物理上是可行的(即不与环境发生碰撞)。只有从这个缩小的解决方案空间中,物理双胞胎才能寻找最快的组装计划。实验表明,该系统确实收敛到最快的装配计划。此外,与直接在物理机器人上学习相比,模拟中的预训练大大减少了收敛前的交互次数。

更新日期:2021-10-17
down
wechat
bug