A system including a memory having instructions causing a processor to perform operations, for planning a grasping-device approach to an object by a grasping device, a pre-grasp device position, and a pre-grasp device pose. The operations comprise obtaining input data including grasping-device data, object data, and environmental-constraint data, determining, based on the grasping-device data, a grasp volume model, determining a test approach vector, and determining, using the vector, the constraint data, and the model, whether approach vector modification is needed. The operations also include modifying, if modification is needed, the test approach, yielding a resulting approach vector, and determining, if modification is not needed, that the test approach is the resulting approach vector. And the operations include determining a virtual floor indicating a position below which the grasping device cannot move, and determining, based on the resulting vector, virtual floor, model, and object data, the device approach, position, and pose.