The problem of friction compensation at very low velocities is investigated experimentally in this thesis. Two recently published friction compensation strategies, based on two well-known friction models, are evaluated experimentally on a direct drive robot arm. One is an adaptive nonlinear dynamic friction compensation method using the LuGre friction model; the other a decomposition-based friction compensation method based on a friction model parameter linearization approach. In the first experimental stage, a new velocity estimation method is implemented for velocity measurement, which significantly improves the effectiveness of friction compensation for all tested methods. In the second experimental stage, a high accuracy laser sensor is selected to replace the DDARM’s internal encoder as the source of the position and velocity signals. This laser sensor enables the relationship between friction torque and its corresponding velocities to be determined. The friction compensation experiments have also been repeated to test the efficiency of the determined friction model.