Container multimodal transport faces many uncertainties in practice. To improve operational efficiency and reduce carbon emissions in freight transport, this study develops a multi-objective optimization model for container multimodal routes that incorporates demand and time uncertainties as well as carbon emissions. The proximal policy optimization (PPO) algorithm identifies robust transport paths facing uncertainty and assesses the model’s sensitivity to price fluctuations and carbon tax rates. Empirical results for the Chongqing–Singapore container route demonstrate the strong applicability of the PPO algorithm. Compared with traditional routing methods, the algorithm yields a lower late-arrival rate and delivers clear advantages in risk avoidance and cost control, thereby effectively reducing carbon emissions in line with carbon-reduction policies and offering practical guidance for logistics firms. The model operates under the assumptions of indivisible cargo and single-visit constraints at nodes, which impose certain limitations. In addition, the current model requires substantial computational resources, which may limit its applicability for smaller companies. With continued optimization, however, the approach advances the industry toward data-driven, intelligent decision-making.