Pneumatic grippers are hugely employed in robotic applications. Nonetheless, their control is not easy due to difficulty in managing the pressure inside their air chambers. Pneumatic grippers have often simple structure though the lack of affordable control algorithms complicates their usage. Motivated by these reasons, we wish to deliver a new control architecture for the closed-loop control of pneumatic grippers actuated by pressure regulators. The proposed architecture is composed of a main controller resorting to an optimization algorithm and of a state observer that estimates pressures in both gripper chambers, along with the exerted force. Instead, measured quantities (i.e. physical pressure in the gripper chambers and force recorded by a load cell between the gripper fingers) are used as inputs for the state observer to improve its output. The pneumatic gripper performance benefits from the joint action of the controller and of the state observer, as experimentally demonstrated. The gripper response will be shown for different types of inputs and on different setups.