Experimental Results of Safe Model Reference Adaptive Control With Performance Guarantees

The safety of a dynamical system is defined by requiring its trajectories to remain within a predefined safe set for all time, a condition known as the forward invariance of that set. The Model Reference Adaptive Control (MRAC) framework adaptively updates controller parameters in real time to ensure that an uncertain system follows the desired dynamics of a reference model, maintaining performance under uncertainty. By integrating Control Barrier Functions (CBFs) into the reference model, one can generate reference commands to keep the reference states within predefined safe regions, forming a safe reference model. In addition, strict performance guarantees on reference tracking constrain the system to remain within a bounded distance from the safe reference trajectory, thereby preserving safety. This work presents experimental results demonstrating a set-theoretic MRAC architecture with a safe reference model that ensures bounded tracking error and maintains system safety through strict performance enforcement, validated on Quanser’s Cube servo system.