GRATP Model Based on Comprehensive Training Cost: Solving Collaboration Problems in Real-World Scenarios

As an extension of group role assignment (GRA), GRA with a training plan (GRATP) is proposed to find the optimal assignment and training plan by maximizing the total benefit. The training plan indicates the training program of each agent, thereby enhancing its corresponding role-playing abilities and further affecting the role assignment. However, only the downtime loss is considered in the existing GRATP models, while the cost of training program is ignored. Moreover, the capability improvement brought about by training is related to the agent’s familiarity with the role, which is also neglected by the existing GRATP models. Therefore, we formulate the training-related role assignment problem while taking into account the comprehensive training cost, which is composed of the downtime loss and the cost of training program. In the formalized problem, different training programs are attached with different costs. Specifically, the cost of a training program is related to the weight and the team’s performance in the corresponding role. In addition, the capability improvement function is set to be positively correlated with the agent’s familiarity with the role. Furthermore, the agent will not be trained if its initial ability value exceeds a certain threshold, which is in line with actual scenarios. The effectiveness of the proposed model is verified by experiments.