Sorting of heterogeneous units is a self-organized behavior which is seen in many biological systems. One of the best examples of such systems is a system of biological cells of heterogeneous types that has the ability to self-organize into specific formations, form different types of organs and, ultimately, develop into a living organism. Earlier research in this area has indicated that such self-sorting behaviors in biological cells and tissues are made possible because of difference in the adhesivity between different types of cells or tissues. Inspired by this differential adhesivity model, this paper presents a decentralized approach based on differential artificial potential to achieve the self sorting behavior in a swarm of heterogeneous robotic agents. The method is based on the proposition that agents of different types experience different magnitude of potential while they are interacting with agents of different types. An analysis of the system with the proposed approach in Lyapunov sense is carried out for stability. Extensive simulation studies and numerical analysis suggest that the proposed method would always lead a population of heterogeneous agents closer to the sorted or segregated configuration.