In order to function safely in a dynamic environment with humans and obstacles, robots require active compliance control with force feedback. In these applications the control law typically includes full dynamics compensation to decouple the joints and cancel out nonlinearities, for which a high-fidelity model of the robot is required. In the case of a 3D-printed robot, components cannot be easily modeled due non-uniform densities, inconsistencies among the 3D printers used in manufacturing, and the use of different plastics with mechanical properties that are not widely known. To address this issue, this thesis presents an adaptive control framework which modifies the model parameters online in order to achieve satisfactory tracking performance. The inertial properties are estimated by adapting with respect to functions of the unknown parameters. This is achieved by rewriting the robot dynamics equations as the product of a matrix of known nonlinear functions of the joint states and a vector of constant unknowns. The result is a nonlinear system linearly parameterized in terms of the of the unknowns, which can be estimated using adaptation laws derived from Lyapunov stability theory. The proposed control system consists of an outer-loop impedance controller to regulate deviations from the nominal trajectory in the presence of disturbances, and an inner-loop force controller to track the joint torques commanded by the outer-loop. The proposed system is evaluated on an early prototype consisting of a 3DOF leg, and two actuator test setups for the low-level controller.