Obstacle avoidance pathfinding

To properly implement obstacle avoidance I see two possible options:

Ray cast

http://www.red3d.com/cwr/steer/Containment.html

Advantages:

• Very simple implementation (similar to existing channels)

• Relatively cheap to include moving obstacles. Might not be that helpful though as a moving obstacle is likely to be small and so using sound with predictive steering will produce better results.

Disadvantages:

• Concave obstacles could cause agents to get stuck.

• More effort by the user to use this data.

Navigation mesh

http://gamedevevacademic.tumblr.com/post/69849130309/ai-pathing

https://github.com/recastnavigation/recastnavigation

Advantages:

• Obstacle avoidance and path finding come together.

• Avoiding complex shapes is done "intelligently". ie won't get stuck in concave obstacles.

• The process of creating navigation meshes could be automated.

• High poly obstacles have the same impact on runtime as low poly.

Disadvantages:

• Requires creation of navigation mesh (which, at least to begin with, would be the responsibility of the user).