Many engineering problems that involve hierarchical control applications, such as demand side ancillary service provision to the power grid, can be posed as a robust tracking commitment problem. In this setting, the lower-level controller commits a set of possible reference trajectories over a finite horizon to an external entity in exchange for a reward corresponding to the size of the reference set and the allowed margin of tracking error. If the commitment is accepted, the lower-level system is required to track any reference trajectory that can be sampled from the committed set. This paper presents the framework of robust tracking commitment and a method to solve the optimal commitment problem for constrained linear systems subject to uncertain disturbance and reference signals. The proposed method allows tractable computations via convex optimization for conic representable uncertainty sets and lends itself to distributed solution methods. We demonstrate the proposed method in a simulation based case study with a commercial building that offers frequency regulation service to the power grid.