How can I Find Walking Paths for Different People in a Graph With Reinforcement Learning?

Question

I don't know it is possible or not with reinforcement learning but my question is about finding walking paths for different people in a graph. A sample image that I draw to help the problem description is here:

The problem is that: There is a garden with graph representation. The black lines are walking ways on the garden and the reds are vertices, so junction points of ways. Blues are people and they can walk over ways and they can only change their ways on the vertices(red points). There are no limit for people, there can be 1 or 10+ people in the garden in same time. People wants to reach to the green gate with the shortest path but while they do that, they must not meet to each other in the same vertex in the same time. For example, person A can be in vertex1 and person B can be in the vertex2, but they both can be in the vertex3 together. I draw example paths for each people with pink color, so my purpose is finding path for each people in the garden which they reach to the green gate without meeting any of them with each other. It is like a time series problem because each person can walk to a vertex in a unit time and all people's speeds are equals.

Is there a similar problem or solution for this? Or what can I do to solve this problem?

Answer 1

Not a solution but for an advice, https://gym.openai.com/ has similar problems and also their solutions. I used once q-learning of basic implementation on https://gym.openai.com/envs/FrozenLake-v0/ this environment but on this environment works with one agent(people). You might want to check this out. Furthermore look at ant-colony and particle swarm optimization algorithms.