Optimal Path Algorithm For Earthquake Relief Based On Q-learning

Earthquake is very dangerous. Strong earthquake will make the building collapsed on both sides of the road and the some roads are blocked. Earthquake disasters often lead to the destruction of the urban road network. Rescue workers outside the city need to send a lot of relief supplies to the disaster area in the first time, and people within the city need as soon as the injured people taken to safe areas. Establishing relief channel with the fastest speed is crucial to save the injured people and reduce property. But, some infrastructure within the city due to the destruction of the earthquake, make the electric power, communication and other system parts and even complete failure, to the disaster area and the outside world cannot communicate information in a short time, which makes the rescue work more difficult.This paper establishes optimal path algorithm of earthquake relief based on Q-learning. It based on artificial intelligence theory and the actual characteristics of urban road network after the quake and a rescue team is regarded as an agent. The environment is urban road network after earthquake. The each node of road network looks as the state of the rescue team. Rescue teams from the node to neighboring node looks as an action. Reliability of the road looks as the return value. Rescue teams after a certain number of learning can get each state-action of the cumulative discount road reliability. It is Q. According to the Q value, the rescue team determines the optimal strategy of selecting the action. With this strategy, rescue team can find the optimal path. In this study, the feasibility of the road network structure is not the rule, Chaoyang district, Changchun City, part of the road network as an example calculated have proved the algorithm. The sensitivity of the model parameters is analyzed. The results show that when the learning rate increases, while the other parameters remain unchanged, agent learning speed; when the discount rate increases, the other parameters remain unchanged, the agent learning speed decreases.At the same time, taking into account that there are multiple rescue teams after the earthquake. The author establishes optimal path algorithm for earthquake relief based on Multi-agent collaboration of communication. After the earthquake damage to the road information as a sharing of information, the rescue team to accelerate the exchange of information through mutual learning speed. Since the initial stages after the earthquake, communication facilities were damaged, with little communication between rescue teams, or no. After a certain time, the communication rate will gradually increase. In this model the communication rate between the rescue team initially zero, as time progresses gradually increase. In order to verify the feasibility of the algorithm, the same part of the Chaoyang district, Changchun City, the road network for instance calculated and analyzed, the results showed that the path optimization model under conditions of multi-agent-based communications and collaboration can find a faster and more rescue learning speed path, multi-agent learning speed than single-agent increased by nearly one-third.