been working on an improved method for detecting navigation obstacles, think I'm onto something now! this costs 1 raycast per frame and maintains a 'radar' around the boat that I can use to adjust steering on the fly

So far I’ve made it without cheating in the AI, they have the same physics as you The next step might be to do some subtle physics nudges if a collision looks imminent and isn’t desired (pirates colliding with eachother is fun, your fleet not so much)

Some extra context for the video: The green spheres show the extent of the radar, red is where it’s hit something Adjusted direction is the white line/sphere The spinning thing is where radar is currently looking The green line is the desired path as given by the navmesh

Seeing a lot of the same questions: - I do use navmesh as well, however the boats will sometimes end up off course from the A* path, in addition to dynamic things like other boats or other obstacles not on the nav mesh

- The 360 scan vs scanning just ahead is nice to have because it gives the AI more information about where it can safely turn when obstacles arise - I said it costs 1 raycast per frame, but really it’s more like X raycasts per second, to account for low frame rates

This is from my recently released game, Sail Forth! Check it out if this caught your eye

Maybe try an egg shape with the boat placed where I put the X. You’ll have a range behind about the same size as it is now, or a little smaller, and a forward that’s much more predictive. It detected it was doing to hit that rock rather late. A further view would fix that.

Cool idea!!

Game dev is a wild and unruly beast and we love to see it ✨

That's clever and I love the use of the gizmos. If it's frame-dependent would lower end devices get lower quality detection? My inclination would be to go by time, say every 20ms or something.

This is possible yeah, I could try to do X number of raycasts per second or something along those lines