Dong

So, there are some place to store the shape data of the invaders. Here the shape data means the pixels which record how the invader should look like in the screen. Here, what your mean is that the TIA will have some registers to record the ovelap event between bullets and the invader. The programmer can get that infomation and then decide what to do to react to this ovelapping thing. Am I correct?

Thanks a lot. It's very helpful. And does atari 2600 have some space other than the 128 bytes for storing the shape data of the invaders? And Could you give me more information about the TIA collision registers? I am not a game programmer. In my mind, I think the collision was detected by some conditions judgement, such as the location overlap. Thanks for your detailed reply.

Dear all, I am confused of how to manage multiple objects in one games? Here we talk aboout Atari 2600. For game, space invader, it has 30 enemies, and one agent. I am very curious about how to manage those objects in Just 128 byte RAM. It seems each one should take two bytes to record the positions(X, Y). But I did't see many changes of ram in stella when the enemies are moving. How to implement that? Could anyone tell me the logic of that? Thanks in advance.

Supposed now we in Breakout Game, we toke an action like moving the paddle right. Then the the "game engine" presents the next frame (in which the ball is moving to next position, the paddle move right). This is deterministic, you give the "game engine" the action to take, then it present the next frame which is only determined by the last frame(or the state) and the action. But in some games, even we know the current state and which action to take, but the next state could have several options. For example, in the Breakout game, if we have state S, and action A to take, then the next state must be S1 based on S and A. But some other games could be: state S and action A, but the next state could be 70% for S1 and 30% for S1'. So my question is, are atari 2600 games like the first example or the second example?

By deterministic, I mean, if I toke an action based on current state which is the frame presented on the screen, is the next frame deterministic? Is there some possiblity control of the next frame, such like 70% for Frame_next, 30% for Frame_next_1?

Dear All, you guys are really nice. I am truely sorry to not reply in time. I didn't recieve the notification of this website. After reading all your replies, I have some new questions. Even we have 128 bytes, the games still run very well. When we allocate some bytes for the games, this bytes store some really important information, such like in Pong game, the locations of ball and two board .Does those bytes will be moved to other place or temporily reclaimed when the game are running? I think it will not do that. For the important bytes in the ram, when it was "born" from the start, it will be reclaimed when the game is over. So if we took the whole ram or we extract the important bytes(by somehow, we suppose we can do it) as the state, it should be right because the important information are stored in the ram during the whole time. Am I correct? Thanks in advance.

Thanks for your detailed reply. Appreciate it.

Thanks for your reply. Is the jitter/roll caused by the mismatch of clock? t seems that every instruction will cost some clock-circle to exectute. The programmer need to sync it with the visualization process. Am I right? So if we don't want the visualization, we can remove it safely without affecting the real game logic computation. By the way, does visualization cost the majority time of the whole CPU time? I haven't read the source code of atari games. It seems that the source code is hard to find. Thanks.

Hi, guys. I have some questions about the programming of atari games. Here is my understanding of atari programming. Logically, a game should contains two parts. First part is the logic computation which is the internal game logic(such like computing the rewards, relative locations and so on.) The second part is visualization to player. This part is time-comsuming (which is to scan all the scanlines). First question: Am I right? Second question: If I am right, Can I remove the part of scaning all the scanlines if we don't want to visualization(which time-consuming)? I heard the people in the forums are very nice. Thanks in advance.

Now I am working on how to represent the state of an Atari game for Reinforcement Learning. I read some papers that using the whole Ram as the state. But the wikipedia of Atari 2600 said that the 128byte could inculde three parts which are scratch space, call stack and the state of the environment. I am wonderring if this is right. And If it is, how to extract the state of the environment?