In the automated operation of the chemical laboratory, using a robotic arm to pick up volumetric flasks is a core step of the operation process. By implementing reasonable path planning, the grasping operation of the robotic arm can be made efficient and precise. In this scenario, the traditional Rapidly-exploring Random Tree Star (RRT*) algorithm suffers from low sampling efficiency and numerous sharp path turns. To address these problems, this paper proposes the Vision-guided Multi-constraint RRT* (VM-RRT*) algorithm, which integrates visual guidance sampling and multi-constraint paths. Firstly, the algorithm determines the spatial coordinates of the volumetric flask through target detection, reducing ineffective exploration and accelerating path convergence. Subsequently, it uses cubic B-splines to fit the path and improves the density of data points through spline interpolation methods. Combined with low-pass filtering, it further reduces noise and eliminates sudden changes in the end-effector speed and acceleration of the robotic arm, realizing multi-constraint trajectory optimization. Experimental simulation results show that the average planning time of the VM-RRT* algorithm is 3.27 seconds, which is approximately 20% shorter than that of the traditional RRT* algorithm (4.07 seconds), effectively improving the experimental efficiency. At the same time, the end motion parameters of the robotic arm are effectively controlled, providing support for laboratory automation.
Yuan et al. (2026) studied this question.