Kinematic Optimization of a Robotic Arm for Automation Tasks with Human Demonstration

TL;DR

The paper tackles task-specific robotic arm design by learning from expert demonstrations of full-arm motion to reduce over-engineered solutions. It defines an optimization over Denavit-Hartenberg parameters, with forward kinematics $A_{ee}(q, \Phi)=\prod_{i=1}^n A_i(q_i,\phi_i)$, to minimize $f(\Phi)=\frac{1}{T}\sum_{t=0}^T G_{\Phi,t}$ plus a penalty $E(\Phi)$. A novel RA-PSO variant of particle swarm optimization guides the search in a large, non-linear design space, with local exploitation inside a feasible subset $\Omega_f$ and a split of angular and length variables updated at frequency $D$. Experiments on two manufacturing tasks show the method yields minimal-DOF arms (3-DOF and 5-DOF) that track recorded paths and obstacles, reducing computational effort and enabling rapid, modular hardware deployment.

Abstract

Robotic arms are highly common in various automation processes such as manufacturing lines. However, these highly capable robots are usually degraded to simple repetitive tasks such as pick-and-place. On the other hand, designing an optimal robot for one specific task consumes large resources of engineering time and costs. In this paper, we propose a novel concept for optimizing the fitness of a robotic arm to perform a specific task based on human demonstration. Fitness of a robot arm is a measure of its ability to follow recorded human arm and hand paths. The optimization is conducted using a modified variant of the Particle Swarm Optimization for the robot design problem. In the proposed approach, we generate an optimal robot design along with the required path to complete the task. The approach could reduce the time-to-market of robotic arms and enable the standardization of modular robotic parts. Novice users could easily apply a minimal robot arm to various tasks. Two test cases of common manufacturing tasks are presented yielding optimal designs and reduced computational effort by up to 92%.

Figures (6)

• Figure 1: Illustration of a robot kinematics evaluated on a recorded task. In this example, the robot end-effector must follow a recorded path of the human hand (circular red markers) while its arm should follow the human arm markers (circular blue and green markers) to avoid obstacles. The dashed curves are the paths of the optimal robotic arm that best track the recorded human ones.
• Figure 2: (a) Illustration of a robot parameterized by $\Phi$ where any point along it can be represented by $\mathbf{s}_\Phi(\sigma,\mathbf{q})$. Points $\sigma=0$ and $\sigma=1$ are the base and end-effector points, respectively. (b) Area to minimize (yellow) for a better fit between the robot arm (solid lines) and lines (dashed) formed by the recorded markers (red).
• Figure 3: Recording of a human arm path in an industrial (Scenario I) packaging and (Scenario II) welding. A robotic arm is to be optimized to accurately track the recorded path. The end-effector of the robot must follow the path of the hand markers while its arm should follow the human arm markers to avoid obstacles. For Scenario II, the EE must also track the orientation of the human hand.
• Figure 4: Fitness results for Scenarios (left) I and (right) II.
• Figure 5: A near-optimal 3-DOF robotic arm tracking the human recorded paths and performing the pick-and-place task.