This paper introduces a robust distributed model predictive control (DMPC) strategy for constrained continuous-time nonlinear systems coupled through their cost functions. In the proposed technique, all the subsystems receive the assumed control trajectories of their neighbors and compute their controls by optimizing local cost functions with coupling terms. Provided that the initial state is feasible and the disturbances are bounded, a two-layer invariant sets-based controller design ensures robustness while appropriate tuning of the design parameters guarantees recursive feasibility. This paper first derives sufficient conditions for the convergence of all the subsystem states to a robust positive invariant set. Then, it exploits the κ ∘ δ controllability set to propose a less conservative robust model predictive control (MPC) strategy that permits the adoption of a shorter prediction horizon and tolerates larger disturbances. A numerical example illustrates that the designed algorithm leads to stronger cooperation among subsystems compared to an existing robust DMPC technique.