School of Mechanical Engineering Seminar
Wednesday, November 2, 2022 at 14:00
Wolfson Building of Mechanical Engineering, Room 206
Optimal Configuration of a Robotic Manipulator to Perform
Inbar Meir Ben-David
Dr. Avishai Sintov, School of Mechanical Engineering
and
Prof. Avital Bechar, Volcani Center
Robotic arms are the foundation of modern automation for manufacturing. They accommodate production lines and perform the majority of tasks such as assembly, machining, painting, welding and packaging. However, these highly capable robots are usually degraded to simple repetitive tasks such as pick-and-place or welding along the same course. On the other hand, designing an optimal robot for one specific task consumes large resources of engineering time and costs. Moreover, common methods search for collision free paths for robotic arms. However, these are unlikely to be found in cluttered environments where objects must be cleared in order to reach the goal. Furthermore, robotic manipulators are usually designed for interaction with the environment solely using their end-effector.
This work seeks for the near-optimal robot configuration to perform a specified task based on human demonstration. The proposed method searches for the robot design variables and the robot placement in the world. An optimal robot is the one that incorporates the minimal DOF and provides best accuracy during task execution. In addition, the optimal design is given with a required path to complete the task. The path is composed using a meta-heuristic method in order to find the joint values to perform the task. This approach takes into consideration the entire robot arm (joint and links), to perform tasks in cluttered environments or to avoid obstacles. The proposed method can also be used to plan a robot path along a human demonstration for existing robots. We provide a comparative analysis to identify the most suitable algorithm to solve this optimization problem. Different known methods, such as Artificial Bee Colony, Genetic Algorithm and Simulated Annealing, are tested and compared. Furthermore and to overcome the highly non-linear and non-continuous search space of the problem, a new algorithm is proposed termed Robot Arm - Particle Swarm Optimization (RA-PSO).
Moreover, we define a new evaluation index Normalized Computation Effort Index (NCEI) that combine the convergence iteration and valid particles.
The new method can find the same robot design of the standard method with lower NCEI.
To test and establish our method, we use three diffident test cases. The first is a classic pick and place of an object from a conveyor while avoiding an obstacle. The second scenario is of a robot performing a welding task and requires to track position and orientation of the human demonstrator. The last scenario is a robot moving inside a palm tree canopy in order to perform the dilution task. The palm tree canopy represents a cluttered environment that a collision free path will not be found by the classic path planing methods. In brief, the results show improvement in the NCEI of 90\ ,69%, 53% and 44% in average for all scenarios for n = 3,4,5,6, respectively.
Join Zoom Meetin https://us02web.zoom.us/j/82108132163?pwd=Z2h4UzNzUS9mbXplT0lMU1pZenFEQT09
.png){kind=spacer}
