The global epidemic of COVID-19 has broken the inherent balance of the manufacturing industry. The rapidly changing market demands and the shortage of labor force make it unable to respond to the flexible changes of manufacturing industry in time. Especially for the assembly of complex products, there are problems that robots cannot complete alone, and the efficiency of manual assembly is low. Human-robot cooperation (HRC) combines the advantages of human and robot, which can greatly improve a manufacturing system’s flexibility and efficiency. In this chapter, we take a CPS-oriented approach to HRC assembly (HRCA). The framework of the HRC assembly CPS system is presented, and the various elements of the collaborative assembly environment are described in detail. We also give the representive information for complex assembly processes in the cyber world, and this chapter proposes operation, action and behavior models for HRC. Then, reinforcement learning is used to unfold the HRC decision strategy based on task states. Finally, a case study is implemented, which verifies that efficient human-robot cooperation with CPS will improve the rapid response and robustness of the assembly system from uncertainty.