I''m writing my own 2D RTS game based in Isomatic Tile-based, I interested in how AOK and SC ( even RA2 ) can make units so well that will avoid obstacle and solve the collision event automatically, and perfect. I tried to wrote the such Collision Solving System that records the event when the collision of two units occurred. Then it can solve these two units to avoid the collision, it may let one of them just move away to avoid the path of the other. That what I think. I had studied the Collision Solving System of those game above for a few while. SC has no such system. If I let a marine ( which is the solider of Terran ) stand in the middle of a group of several marines who stand around as cyclic. The marines outside won''t move away to let the one stand in middle out when he want to. let''s focus to the SCV now. when there''re lots SCV extracting resources, they won''t get collision because when it won''t check collision with SCV who is extracting. This way makes it having less collision problem. I felt it''s unreality, but it''ll get lots collision while the worker ( such as SCV ) extracting resources... Now focus to the Group problem. In AOK, who has the best group system in RTS games, I''m not sure if it has Collision Solving System or not. When the group of several ( or more ) soliders are grouping, they won''t get collision neither. This makes less collision occur as the SCV extracting in SC. So I wonder that can''t we check collision as reality in everywhere ? I wouldn''t like to make a game which is unreality , which is just like a Cartoon. BTW, How to solve the collision of two units as well as we can? I meant, what''s the best way to determine who move away and who don''t have to. The collision event are in several case, such like Unit1(Moving) to Unit2(Moving), Unit1(Stop) to Unit2(Moving), Group1(Moving) to Group2(Moving), Group1(Moving) to Group2(Stop) ... etc. How to determine which way to solve these event? Besides, I hope you guys can understand what I''m talking about while I have a poor English ... and, Thanks anyway.