In response to the high complexity, stringent real-time requirements, and intensive human-machine interactions characteristic of collaborative tasks among crew members across different platforms in intelligent environments, this study proposes a crew attention allocation model that integrates information processing procedures with platform collaboration factors. By constructing a task network model and extending the Team-GOMS (Goals, Operators, Methods, Selection rules) framework, the model incorporates critical path analysis, social network modelling, and quantitative methods for team situational awareness to establish a multi-level predictive system for attention allocation. In typical multi-vehicle collaborative task scenarios, this model can predict task completion times and effectively identify critical bottleneck paths, thereby providing theoretical support and quantitative tools for optimising human-machine collaboration efficiency in complex task environments.