A constrained robot manipulator with contact friction between its end-effector and environment is considered. A new intelligent hybrid position/force controller is designed. The controller includes two major parts. The first, part which is called the main controller consists of two closed loops corresponding to motion tracking and force tracking objectives. In each loop the decoupled governing equations of the system are initially linearized, using a feedback linearization approach, then the linearized loop is controlled by a linear controller. The second part, which is called the tuning controller, is an adaptive neural network (NN) controller to compensate the model based deficiencies of the first part. The main contribution of this paper is to improve the model based controller, by use of the neural network controller in the presence of some sort of uncertainties in the system modeling. The stability condition of the closed loop system is approved by using the Lyapunov, passivity and Linear Matrices Inequalities theorems. The performance of the modified controller is simulated for a two-link robot manipulator which interacts with a horizontal surface. The results show an excellent enhancement in control strategy by using neural network controller. The new controller is implemented on a real two-link robot manipulator experimentally. In order to accomplish this implementation, first the LuGre parameters are identified by a separate setup which was built to emulate the end-effector and surface friction. Then parameters are then utilized in the manipulator controller. Experimental showed a very outstanding performance for the added NN controller in the control of the system especially in the end-effector position control. Keywords: Constrained Robot Manipulator, Hybrid Force and Position Control, Linear Matrix inequality (LMI), Neural Network