The estimation of external forces exerted on a robotic manipulator with harmonic drive gearing without a force/torque sensor is considered. Manipulator dynamics, together with motor current feedback are used to estimate external joint torques, which are transformed into estimated external end effector forces using knowledge of the manipulator's kinematics. Adaptive control is used to tune the parameters of the robot's modeled dynamics, while adaptive radial basis function (RBF) neural networks are used to learn the friction dynamics. Admittance control without force sensing is attempted on a two degree of freedom manipulator. Readings from a six-axis force/torque sensor mounted on the manipulator are used to validate the force estimates during the estimation phase.