Addressing the dual challenges of efficiency and risk resilience in large-scale heterogeneous resource scheduling, this study constructs an integrated computational decision-making framework combining trend forecasting, dynamic pro-gramming, and stochastic simulation. The research first proposes a five-dimensional multi-model prediction framework, employing non-stationary programming algorithms such as logistic regression, Grey prediction, and learning curves to perform parameter identification for core variables including future logistics scale, costs, and system reliability. For largescale cargo transshipment tasks, a Time-Space Extended Linear Programming (TELP) model is developed. Through a two-level optimization strategy combining outer-layer time enumeration and inner-layer linear programming, the Pareto frontier between transportation duration and economic cost is successfully decomposed, identifying optimal equilibrium solutions under heterogeneous collaboration modes. To address random disturbances during system operation, a robustness assessment mechanism based on Conditional Value at Risk (CVaR) is introduced. Combined with Monte Carlo simulation, this quantifies the impact of tail risks on schedule duration. Finally, by designing an adaptive material classification and allocation algorithm alongside a reliability-aware inventory optimization model, the expected value loss of the system is reduced by 67.6% under service level constraints, while optimizing the environmental load caused by emergency dispatching. The universal computational paradigm established in this study provides precise mathematical tools for stability evaluation and automated scheduling of complex systems under extreme conditions.

spatio-temporal extended linear programming, five-dimensional forecasting framework, conditional value at risk