Imagenet large scale visual recognition challenge (ILSVRC) has increased the image recognition ability based on machine learning including Deep Neural Network, whereas Amazon Picking or Robotics Challenge has demonstrated that the image recognition methods are useful for robots to pick and place many types of objects. Pick and place does not require accurate motion control, and the next challenge thus should involve the requirements for high motion accuracy. In this context, World Robot Summit industry category (WRS) was held to pursue the robot ability for completing a complex assembly task. However, even top teams did complete only several parts of the assembly tasks. It indicates a large gap between robots and human at assembling ability. It also indicates the current limitations and potentials of robotic assembly technologies.
Peg-in-hole, screwing, and positioning requires high accuracy control of object’s position and pose as well as elasticity for reducing the contact impact between objects and compensating the error in alignment. Unfortunately, even advanced image based recognition methods cannot provide enough accurate recognition for completing such assembly tasks. Vision sensor and simple parallel or vacuuming gripper are not enough for completing assembly tasks. The realization of autonomous robotic assembly requires a wide variety of elemental technologies including the development of multi-purpose endeffectors for assembly, embedding of mechanical compliance for reducing contact impact, the utilization of tactile and force-torque sensors facilitating operations, high qualified calibration for precise motion generation, alignment and positioning functions, error recovery techniques, and parameter tuning. Furthermore, system integration technique connecting these technologies is essential for working the whole robot system. Which kind of assembly tasks are targeted is also important key because the target selection leads to the acceleration of the creation of new technologies. If focusing on the technologies at the teams who challenged the WRS competition, the accuracy of teaching methodologies and agility at the tuning of the parameters and modification of programs made a big influence on the competition scores. These manual operations revealed the difficulty of the atomization of robotic assembly. A breakthrough is required for autonomous robotic assembly.
This workshop aims at clarifying the current limits and potentials at robotic assembly by investigating not only the assembly challenge program results but also the state-of-art technologies for robotic assembly, to accelerate the generation of new methodologies, strategies, and techniques for automatizing robotic assembly.