Robotic manipulation of deformable objects using vision and touch sensing

迄今为止，机器人操作的研究主要集中在处理刚性物体上。然而，许多重要的应用领域需要操纵非刚性或可变形的物体，如织物，软管和电缆。这样的物体更难处理，因为它们可以表现出更大的行为多样性。机器人操作已广泛应用于工业中的任务，如仓库操作和装配。然而，处理灵活的物体仍然是目前的机器人平台的挑战。这个博士项目将研究使用一个系统，包括机器人机械手配备触觉传感器和立体摄像机的织物操作。深度强化学习将用于改善机器人的操作性能，通过与模拟环境中的物体交互实现在线学习。通过施加挤压，滑动，抓取和移动等程序，可以估计织物的属性和状态。仿真环境中的优化模型将在真实的机器人平台上进行部署和测试。该项目将为仓库运营和灵活材料的组装提供解决方案。

Research on robotic manipulation has mainly focused on handling rigid objects so far. However, many important application domains require manipulating non-rigid or deformable objects, such as fabrics, hoses and cables. Such objects are far more challenging to handle, as they can exhibit a much greater diversity of behaviours. Robotic manipulation has been widely applied in industry for tasks such as warehouse operation and assembly. However, handling flexible objects is still challenging for current robotic platforms.This PhD project will investigate the manipulation of fabrics using a system comprised of robot manipulators equipped with tactile sensors and stereo cameras. Deep reinforcement learning will be used to improve the manipulation performance of the robot, enabling online learning through interaction with objects in a simulated environment.By exerting procedures like squeezing, sliding, grasping and moving, the properties and states of the fabrics can be estimated. The optimized models in the simulation environment will then be deployed and tested on the real robot platforms. The project will create solutions to warehouse operations and assembly of flexible materials.