2D Examples.

[pjohalloran]

• Moved existing examples into a 3D folder.
• Added basic and intermediate 2D examples.
• Added 2D version of the obstacle advanced example.

[pjohalloran]

I am having a bit of trouble with the neighbour set of examples. I wonder @ricardojmendez if you have time, would you mind giving me some advice on how to improve the configuration of them to correctly demonstrate alignment, cohesion and seperation?

[ricardojmendez]

Hi @pjohalloran,

A couple notes after a quick review:

• There's a 3D folder inside the 3D folder.
• It looks like the obstacles on the 3D examples have changed. You probably replaced the prefab by mistake when doing the 2D example.
• For 2D, you may want to have a boid that more obviously change direction - for instance, adding a red arrow on the sprite where the front is supposed to be. This will help with debugging.
• It also look like the tether is too short, which is causing the clustering around the center.
• Finally, they are missing a SteerForSeparation behavior, which is a fundamental part of the flocking behavior. That's why you're getting such tight groups.

Overall, I'd relax the values a bit. Give them a wider detection radius, play with the min/max values on neighbor groups, and relax the tether. Look at the 3D example for the some proportions to give you an idea.

[pjohalloran]

Thank you for the review. I am a little busy in my job at the moment so will not be able to sort this out until next weekend. 👍

[ricardojmendez]

Sounds good, thanks for working on them. Cheers!

[pjohalloran]

Hi @ricardojmendez. I think I have addressed the issues you raised above.

The main issue i was having with the Neighbor examples was that I had the Radar2D detection layer set to 'Nothing'. Fixed that, plus took on board your advice above.

Would you mind having another review? Maybe we are nearly at a point where this can be merged in? Thanks!

[ricardojmendez]

Thanks. Looking at the obstacle avoidance scene, the boids seem a big groggy and keep turning around - I'll take a look at it.

[ricardojmendez]

OK, this is a good example of how tweaking the values end up creating completely different behavior. There's several things that are contributing to this jittery way the vehicles are reacting to each other.

• Spherical obstacle avoidance weight is too high - at 100, it'll override any other behavior when it kicks in. I'd say decrease to 4 or 6.
• The DetectionRadius on the radar is too small. The agents are barely looking beyond their own radius. I'd say increase from 1 to 8.
• MinRadius and MaxRadius on SteerForNeighborGroup are too tight. The MinRadius is just about the same as the vehicle's own Radius, meaning that for a vehicle to definitely be considered a neighbor, it'd have to be almost on top of the current one. The MaxRadius is just 2x the vehicle's own, meaning anything beyond their own diameter is excluded. Try 2/6.
• You may also want to increase the angle to something like 90 degrees, to let them widen the list of vehicles they consider neighbors.
• The RigidBody is not Kinematic. Is that intentional?
• MinDistance on Cohesion should be larger than the boid's own radius - we don't want to apply cohesion otherwise.
• There's still no SteerForSeparation. Try adding one with weight 2, Min/Max of 0/2, Comfort distance of 1, and a multiplier inside the comfort distance of 3.
• Play with the acceleration/deceleration rates. I'd say they're too high, and 1/0.75 would be a good place to start.
• Increase the MaxForce on the vehicle. Try 8.

When in doubt, click "Draw Neighbors" on SteerForNeighborGroup2D - that'll give you debug lines showing who each vehicle considers a neighbor.

I'm attaching a prefab with these new settings.

ObstacleAvoidanceBoid.prefab.zip

[pjohalloran]

Thanks @ricardojmendez. Have updated pull request with your prefab.

Rigidbody should be kinematic, that was not intentional.

[pjohalloran]

Do you think this is good to go now? Or would you like me to improve any other example?

[ricardojmendez]

Hi,

I think it's good to go, thanks! Will merge over the weekend when I'm back on my desk.

Cheers,

-- Ricardo

On 28 Jan 2016, at 04:27, PJ O Halloran notifications@github.com wrote:

Do you think this is good to go now? Or would you like me to improve any other example?

—
Reply to this email directly or view it on GitHub.

[pjohalloran]

Fantastic! Thank you for great review.