With the usage of a pay-per-use business model, cloud computing gives clients Infrastructure as a Service, Software as a Service, and Platform as a Service to provide compute, network, and storage capabilities. Wireless communication may frequently occur in cloud computing to ensure smoot data transmission for storage and processing purposes. In view of the characteristics of frequent changes in traffic speed and many alternative paths in urban areas, the traditional algorithm cannot effectively converge due to a large amount of calculation when selecting a path, and the search efficiency in 3D path planning is low, and it is informal to collapse into stagnation and local optimal problems. Machine learning techniques, in particular, are the driving force behind the cloud backend for emerging paradigms. By resolving issues with job scheduling, resource provisioning and allocation, offloading, load balancing, migration of Virtual Machines (VMs), VM mapping, workload prediction, energy optimization, device monitoring, etc., and these learning techniques essentially improve how these paradigms are used. In this paper, we suggest a process and mechanism to enhance the path selection process that can be utilized in emerging wireless cloud computing environments. To increase the search efficiency in the early stages of the algorithm, the initial pheromone is first dispersed unevenly based on the traits of the best path. Secondly, a dual-objective mixed-integer programming model is established, and dynamically adjusting the values of the weight factors a and of the pheromone and heuristic function accelerates the convergence of the algorithm while taking into account the minimum total distribution cost and maximisation of overall customer satisfaction. The local optimum of the algorithm is reached. The simulation results demonstrate the utility and reference value of the proposed model and enhanced algorithm for planning the distribution of vehicles on complex roads.