Objectives  With the development of urbanization, urban structure and traffic conditions are becoming more and more complex, and it is difficult to design and implement of urban emergency evacuation. Based on existing research results, we propose a novel organized evacuation plans with collaborative optimization strategy, named OrCOS.

  Methods  The object of the OrCOS is the dense crowd in the city which needs to be evacuated immediately. With some strategies about collaborative exits allocating and path planning, OrCOS can reduce the evacuation time and the congestion time of evacuees, and achieve the more efficient evacuation plan. The idea of OrCOS is as follows: (1) Model urban road network and evacuees. Urban road network is defined as vertex‑arc network and residents of one city block are abstracted as an evacuee group with consistent moving features. (2) Determine the shortest path of each evacuee group. Based on the Floyd's shortest path algorithm, the nearest exit and the shortest path of each evacuee group can be determined.(3)Smooth the difference of occupancy time between exits by reallocating exits of specific evacuee groups until the difference is minimized. (4) Re‑plan evacuation route of each group. Because of the interference between evacuee groups, the groups' fastest path need to be replaned to optimize evacuation time.

  Results  Some experiments are carried out in a simulated road network with 247 vertexes and 312 road arcs (449.8 km in total). The total evacuation time (T) and the average congestion time (T_c) are analyzed. Experimental results show that: (1) T and T_c decrease with the increase of the exits' number. When 8 exits are employed, 490 thousand people need 35 min to complete the evacuation. (2) T and T_c increase with the number of evacuees. With 3 exits, 2.44 million evacuees take 4 hours 22 minutes to complete evacuation. In addition, we also conduct comparison experiments between the shortest path evacuation scheme (SE), the staged multi‑exit evacuation scheme(SRE), and OrCOS, to show the differences of T, T_c, and the algorithm running time (T_r), which show that: (1)In most cases, OrCOS often has shorter T than others, especially when the number of evacuees is larger. (2)OrCOS has the slightly better T_c than SRE, and SE has the worst T_c. (3)OrCOS has the worst T_r than others, because of its huge demand for collaborative computing. In summary, OrCOS has stronger ability of evacuation than SE and SRE.

  Conclusions  We propose a novel method: OrCOS, by collaborative optimization strategy, which provide an efficient evacuation plan for urban emergency evacuation. Experimental results demonstrate that OrCOS has better performance in urban road network. It provides a new idea for the multi‑exits simultaneous evacuation in high‑density crowd scene.