Abstract: In order to improve the intelligent equipment level of bioremediation technology, a heavily polluted coke plant was taken as the research environment, and the double deep Q network (DDQN) and ant colony optimization algorithm (ACO) were used to establish a multiple unmanned ground vehicles (multi-UGV) path planning and task assignment system for the topographical features of the coke plant to achieve safe and accurate transportation of contaminated soil in the soil remediation process and improve the efficiency of contaminated soil transportation. The results showed that the multi-UGV transportation system based on DDQN and ACO had good path planning capability, and the ACO task assignment algorithm based on the actual system time cost could achieve a stable reduction of UGV system time cost under different loading quantities compared with other task assignment strategies obtained based on simple linear distance or based on the greedy algorithm.