To minimise traffic congestion and emissions on urban multi-class expressways, five performance indicators were considered, such as total travel spent (TTS), total travel distance (TTD), ramp queuing, exhaust emissions, and fuel consumption. The macro traffic flow model (Multi-class METANET) and emission model (Multi-class VT-macro) were improved. A new high-dimensional multi-objective optimisation algorithm, fuzzy non-dominated sorting genetic algorithm (FNSGA-Ⅲ), was proposed to optimise the on-ramp inflow rate on expressways and the variable speed limit (VSL) on main roads. The new algorithm, FNSGA-Ⅲ, aimed to alleviate traffic congestion, save energy, and reduce emissions simultaneously. This proposed algorithm, which was based on an adaptive network-base fuzzy inference system (ANFIS), could effectively guide the evolution direction of the optimisation algorithm in the iteration process and improve the convergence speed. Based on the actual road network of Guangzhong Road in Shanghai, a simulation was performed. The results showed that the proposed FNSGA-Ⅲ algorithm and the improved multi-class macro traffic flow model could reasonably set the desired speed and ramp control strategy. Hence, compared with the existing single-objective genetic algorithm and multi-objective NSGA-Ⅲ, the proposed algorithm and improved model could effectively alleviate traffic congestion and reduce emissions.