Hi, I'm developing a 2.5D platformer game where the player's goal is to reach finish point first with avoiding obstacles in levels. The control schema of the player is like angry birds(drag and release). I thought that the AI of this game suitible for ML technique. But I'm not an expert of Unity ML Agents, I finished a course on Udemy about Unity ML Agents a year ago. So, I wanted to ask if it is possible or not? What I'm currently thinking; I need only 2 outputs from Tensorflow, the direction vector and the force amount. But I'm not sure what I need to give as an input(observation). I have two ideas; * I can cast about 30 rays 360 degree around the agent and collect the hit distance data of each ray and give this distances as an input as well as the position of the agent. * I can create a render texture that showing the profile of the level and give this render texture to the agent.