To effectively address the requirements of deeply coexisting-cooperative-cognitive (Tri-Co) interactions among humans, robots, and the environment within the context of Industry 4.0, it becomes crucial to conduct evaluation research on Tri-Co capabilities (TCCs) of collaborative robots (cobots), which is a knowledge-intensive task. However, existing research on performance evaluation of cobots has not yet established a common method for constructing a complete performance evaluation index system. The testing methods lack a design basis and fail to evaluate the performance of cobots through operational tasks in industrial production or daily life from the perspective of test tasks. Furthermore, these methods do not facilitate the fusion of multi-source subjective and objective decision-making information, which includes both the knowledge of experts and fundamental parameters of cobots. To this end, this study proposes a multi-source decision-making information fusion framework for evaluating the TCCs of cobots. This framework includes a construction method for an evaluation index system that fuses subjective and objective elements based on statistics, text clustering, and closed-loop feedback mechanism. It also incorporates TCCs test tasks, and an improved fuzzy analytic network process (IFANP). Additionally, it incorporates a combination weighting method that aims to minimise both subjective and objective weights deviations. This framework effectively integrates the subjective knowledge of experts with the objective fundamental parameters of cobots. It accomplishes local TCCs evaluation from the perspective of test tasks. Additionally, it achieves global TCCs evaluation by combining basic performance evaluation indices and the performance of completing the test tasks. A case by Rethink Sawyer (Sawyer) is presented to demonstrate the application process and viability of the developed framework. The experimental results and calculation examples indicate that the research has the potential to provide a certain theoretical basis and reference to solve the performance testing and evaluation issues for other types of robots in engineering applications.
