ML Agents, combine raycast detection with visual observation?

In Unity machine learning, is it possible to combine raycast detection with visual observation, to train the agent?

Specifically I want to use raycast detection to detect fast-moving physics objects like a ball, and other AI agents. I'd also like to use visual observation (like in Snoopy Pop) to teach the agent about the current tile-based level, which would be especially helpful for things like detecting different tile types (floor, wall, ice, lava etc). Ideally I'd like both running simultaneously.

Thank you!