论文标题
无需对象模型的精确选择的自我监督学习
Self-supervised Learning for Precise Pick-and-place without Object Model
论文作者
论文摘要
灵活的选择是机器人技术中的一项基本而又具有挑战性的任务，尤其是由于需要对象模型来进行简单的目标姿势定义。在这项工作中，机器人根据单个，显示的目标状态学习使用平面操纵来拾取对象。我们的主要贡献在于结合原始人的机器人学习，通常由完全跨跨神经网络估算，并与一击模仿学习相结合。因此，我们将地点奖励定义为现实测量和特定于任务的噪声分布之间的对比损失。此外，我们设计的系统以一种自我监督的方式学习，从而实现了最多25000个选择动作的现实世界实验。然后，我们的机器人能够将训练有素的对象放置，平均放置误差为2.7（0.2）mm和2.6（0.8）°。由于我们的方法不需要对象模型，因此机器人能够概括为未知对象，同时保持5.9（1.1）mm和4.1（1.2）°的精度。我们进一步展示了一系列新兴行为：机器人自然学会在存在多种对象类型的情况下选择正确的对象，精确地将对象插入钉子游戏中，从密集的混乱中挑出螺丝，并从单个目标状态中删除多个拾音器动作。
Flexible pick-and-place is a fundamental yet challenging task within robotics, in particular due to the need of an object model for a simple target pose definition. In this work, the robot instead learns to pick-and-place objects using planar manipulation according to a single, demonstrated goal state. Our primary contribution lies within combining robot learning of primitives, commonly estimated by fully-convolutional neural networks, with one-shot imitation learning. Therefore, we define the place reward as a contrastive loss between real-world measurements and a task-specific noise distribution. Furthermore, we design our system to learn in a self-supervised manner, enabling real-world experiments with up to 25000 pick-and-place actions. Then, our robot is able to place trained objects with an average placement error of 2.7 (0.2) mm and 2.6 (0.8)°. As our approach does not require an object model, the robot is able to generalize to unknown objects while keeping a precision of 5.9 (1.1) mm and 4.1 (1.2)°. We further show a range of emerging behaviors: The robot naturally learns to select the correct object in the presence of multiple object types, precisely inserts objects within a peg game, picks screws out of dense clutter, and infers multiple pick-and-place actions from a single goal state.
