The Robotic manipulation in dynamic environments, stable and collision-free manipulation of objects is the most important consideration. In this work, a novel approach is proposed that combines two robust techniques—Dexterity Networks (Dex-Net) for grasping objects and Rapidly Exploring Random Trees (RRT) for path planning. Dex-Net uses an enormous 3D model database to make guesses at stable grasps, and this allows robots to grasp wide numbers of objects, even potentially cluttered objects. RRT then augments this by efficiently searching for safe paths in high-dimensional environments. Together, the algorithms allow for robots to manipulate objects independently within dense environments with confidence. The combined system excels in comparison with isolated methods with 94.7% understanding success, 98% path effectiveness, and 97.5% clutter insensitivity. Such an achievement improves robot system flexibility and dependability. Future research will focus on reducing computational costs, increasing resource usage, and applying the solution to cooperative multi-robot tasks.