As the Development Leader, I updated the trajectory algorithm to realize an autonomous manipulation, which contributes to mitigating the labour shortage in the industrial manufacturing sector. To achieve autonomous control, I successfully solved trajectory planning by introducing the weighted graph theory with a kinematics and dynamics model. This algorithm boosted machine performance to 125% and was patented.
C++, Git, ROS, Robotics Library, Linux OS, Kinematics, Dynamics
This project aims to address the labour shortage in the manufacturing industry by developing an advanced robotic system that enables non-technical workers to operate robots autonomously. This marked Omron company's first venture into the robotics sector.
As the Development Leader, I was responsible for updating the trajectory algorithm for the robot's operation.
I developed a motion algorithm and integrated it into a robotics system, which resulted in a 125% performance of the pick-and-place operation. This patented algorithm would support the broader implementation of robots in manufacturing.
To enable autonomous control, our team addressed three essential requirements:
To satisfy the requirements, our team researched the latest algorithms by attending international conferences and reviewing academic papers, resulting in the identification of a foundational algorithm. In addition, our team utilized the simulator to efficiently evaluate robot performance under various conditions. Furthermore, our team updated the dynamics model to calculate joint torque more precisely using a 3D CAD model.
Although our team did best, the realization of promptness was still a challenging issue. Maximizing machine performance generally takes half a day or more for even an expert robot operator. A shorter path decreases takt time but increases inertia, leading to higher torque that can damage motors. Conversely, a longer path reduces torque but increases the time to reach the endpoint. Optimization needs lots of effort to iterate testing and refinement.
Observing the professional tuning method gave us the breakthrough. Through observation, I could divide the tuning process into two steps: path optimization and motion optimization. After that, I introduced an academic theory, which gained a deep understanding of this solution. I defined this issue as a multi-constraints problem that can be solved by weighted graph theory. To apply this theory, I conducted an experiment to break down a robot’s movement into more than 50 patterns, which fostered practical knowledge about kinematic and dynamic models. Finally, I conducted application tests with the algorithm integrated into robots, successfully meeting requirements in a real-world manipulator.