$Revision: 1.3 $, $Date: 2002/08/31 02:05:44 $ Fri, 30 Aug 2002 13:32:59 > Is the server software that'll be used in the judging available, or do we > have to write our own to test? If the latter, will there be any more > information about how it'll behave - i.e. the distribution of sizes of > map, numbers of packages and money awarded? There will be a server available soon, stay tuned for that. We've been swamped with questions and have been too busy to post it. As for the distribution of various parameters, there will be no more information about that than what's in the description. ----------------------------------------------------------------------- Fri, 30 Aug 2002 13:41:41 > 1) Assume there are no packages on the board. Therefore, no robot can score any points. In such a scenario, how do you determine a winner of the match? Is it the robot that survives the longest? Or, is it the robot with the most money remaining at the end? The number of packages will vary between 1 and 10000. > 2) Does surviving the longest in a match, and saving your money contribute at all to your point total at the end? Saving money doesn't matter for scoring. > 3) Is picking up a package from a location which has no packages a malformed command? No. > 4) Assume we don't have any packages, and there are no packages at our current location. Do we HAVE to move off our location? (We realized we have to bid, but do we have to move?!) No, one can e.g. pick or drop no packages. > 5) How do we submit sample maps? (Inline text or attachments?) Simply attach them in a mail to icfp-judges@cse.ogi.edu. The format for package lists will be documented soon, stay tuned. ----------------------------------------------------------------------- Fri, 30 Aug 2002 14:09:58 > What are the precise rules for when a package is delivered, and what > happens to the package after it is delivered? It's delivered when it's dropped on its destination tile. After the package is dropped, it disappears from the game. Clarification: it only disappears if it's dropped at its destination. If it's dropped somewhere else, it stays where it's dropped. (We are clarifying this in the rules now.) > What happens if the game never ends? It always ends, since every player has to spend money in every turn, and the amount of money is limited. ----------------------------------------------------------------------- Fri, 30 Aug 2002 14:19:52 > In a game with only one robot does it matter how many points the robots > gets in the end? Because he always wins anyway... Always try and maximize the number of points. We've clarified this on the home page. Go back and take a look. > What happens if a robot tries to pick up a package that is not there or > that is too heavy or tries to drop a package that it doesn't have? Does it > die or is it a no-op? Those packages are ignored from the lists, the robot doesn't die. ----------------------------------------------------------------------- Fri, 30 Aug 2002 14:20:53 > It seems to me that it may be desirable for a robot to do nothing for a > turn. For example, not nearing a square that is advacent to both another > robot and a trap. That can be done by e.g. doing a pick command with an empty list ----------------------------------------------------------------------- Fri, 30 Aug 2002 14:41:56 > What happens if a robot tries to go into a wall? Is that a no-op? That is a no-op. > According to the end of subsection "Transmitting the board" north is > downwards. But this is contrary to the example in "Pushing and Bidding" > where 2 moves up and north. The maps are transmitted upside down, as opposed to the examples in the Pushing and Bidding section. > According to the second bullet in the section "The robots" "Packages may > not be picked up [..] if there are no packages available when the robot > gets to executes its command, for example if it got pushed". The question > is: How is the robot supposed to execute a command if it got pushed? True, that cannot happen! We'll clarify. When we do it will be reposted on the challenge page. ----------------------------------------------------------------------- Fri, 30 Aug 2002 14:41:50 > what happens to a package when it is dropped at its destination? does it > just continue to sit there, able to be picked up and delivered by > robots? or does it disappear? It disappeerrs. see Version 2 of the task on the web page. ----------------------------------------------------------------------- Fri, 30 Aug 2002 14:45:27 > Another clarification please. > > "The game ends when either all the packages have been > delivered, or all robots are dead". > > 1. Are clients supposed to detect termination and exit? > Will this be a criterion for judging correctness? no, and no. > 2. If so, how is a client supposed to know when the game > ends? Yes we can detect when all robots are dead, but we > cannot know when all packages have been delivered. see above. > Unless, perhaps, there is exactly one pacakage at each > home base? The specification does not specify, so we have > assumed that there may be many packages at each base. There may be many packages at each home base. there may be also packages on other squares if robots (erroneously) dropped them there and that was not thier deilvery point. ----------------------------------------------------------------------- Fri, 30 Aug 2002 14:48:44 > What is the maximum weight or maximum number of packages a player can carry (or > is it purely dependent on game start parameters) The number is only limited by the weight. the weight is a game parameter and is sent to the clients at the beginning of each game. > Can more than one robot stand on home base at once? If two or more robots try > to get out of a home base the same way, will the lower bid robot get out, with > he rest bumped? robots cannot step on top of each other, they push instead (see rules). they could be at the same location when they start off a new game. ----------------------------------------------------------------------- Fri, 30 Aug 2002 14:52:57 > Since there is no indication of robot death in the server response, > it appears that your intention is that players do not learn of the death > of robots until the turn after the one in which the death occurred (at > which time it can be inferred from the absence of the robot from the > next server response). Correct? That's how it is. ----------------------------------------------------------------------- Fri, 30 Aug 2002 14:58:22 > When a robot dies, does it keep its points? In other words, can the > winner of a given game be a dead robot? Yes, the score a robot gets in a game is the score it has when it dies (or the game ends). ----------------------------------------------------------------------- > Is there an upper bound on board sizes? Yes, the maximum size of the board is 1000x1000 squares. ----------------------------------------------------------------------- Fri, 30 Aug 2002 15:11:11 > From the way the task is described, it seems that robots may be added > serially, is this the case? If so, how will it be determined who which > order the robots enter? When we judge entries, all robots will enter the game at the same time. (However, the protocol allows for robots to join at any time during a game, so we might provide test servers that make use of this.) ----------------------------------------------------------------------- Fri, 30 Aug 2002 15:14:5 > It is entirely possible that a robot fails to move or pick up/drop a package > during a round if the robot ends up getting blocked for some reason. How is this > indicated in a server response line? Will the robots unchanged coordinates simply > be transmitted in the response? The response clearly indicates what happened to the robot, i.e. where/if it moved, what packages were dropped/picked. ----------------------------------------------------------------------- Fri, 30 Aug 2002 15:15:47 > Clarifications please > > 1. What is the range of legal package weights. > Presumably >= 1? What is the max? We haven't decided on a legal range, but presumably >= 1. > 2. Can new robots appear during a game, or > will they all be in the initial configuration? New robots can appear during a game. When we judge entries, all robots will enter the game at the same time. > 3. When the server sends a reply ("Server's reply to commands"): > > a) can the X and Y actions appear anytime other than > in the initial configuration message? Yes. > b) can an X command appear without an immediately following > Y, or a Y without an immediately preceding X? > > What happens if we get an X command for a new robot > without a corresponding Y command or vice versa? The X and Y commands have now been merged into one command, so the Y command will effectively always follow and X command. > 4. In "Server's reply to commands" it is stated that > "the server will send an individualised reply to each player". > But what follows is not player-specfic. Indeed we believe that > on the contrary, the server must send a (semantically) identical > reply to every player. Is that so? Yes, it's the same reply to every player. > Presumably the "servers reply to commands" includes the > actions of the player's own robot? That is implied by the > "robot is listed iff it was alive at the beginning of the turn". > Yell if this is wrong interpretation. > Yes, the player's own robot's actions are included in the reply. ----------------------------------------------------------------------- Fri, 30 Aug 2002 15:18:39 > Don't you think that numbering the southernmost row as "1" > will create lots of additional confusion that could have > been avoided if you had instead numbered the northernmost > row as "1"? > > Your choice results in the maps in the input files > being rendered (in a text window) as upside down. That might cause confusion, yes! But follow the spec as written. ----------------------------------------------------------------------- Fri, 30 Aug 2002 15:20:06 > I am not 100% about the behaviour (or lack thereof) of the packages in the absence > of robot intervention. I think the following assumptions are valid, but are not > stated directly in the problem wording: > 1. Packages are not added during the game (the total available is fixed before any > robots join). True. > 2. Packages do not move of their own accord True. > 3. Delivered packages are no longer available during the game (i.e. > are never reported by the server as available for pick-up). True. > 4. Packages "lost" through robot death are no longer available during > the game (never reported available). True. > Some of these seem obvious, but it's usually better to check :-) Could > you please verify these points? It seems you got everything right! ----------------------------------------------------------------------- Fri, 30 Aug 2002 15:21:42 > we've got a few questions regarding the spec: > > what happens to a packet when it is delivered? does it disappear? > if not, does delivering it again score? It disappears if it's dropped on its destination. If it's dropped somewhere else, it stays on that tile. > does a robot which dies prematurely keep its score? (we assume it > does, but it would be nice if the spec would stress this point) Yes, it keeps its score. > does every home base contain at least one package? It might be empty. ----------------------------------------------------------------------- Fri, 30 Aug 2002 15:22:45 > Do the delivered packages vanish from the world, or can a robot get itself > a lucrative pick-drop-pick-drop-ad-nauseaum deal (since the rules state > that correctly delivered package scores points, and there is no mention of > getting negative points for picking up an already delivered package) The packages vanish if they are delivered to their destination, otherwise they stay where they're dropped. ----------------------------------------------------------------------- Fri, 30 Aug 2002 15:33:37 > How many players can join the server? Many. > "Server: send initial server response (with the initial location of all > players)" > Does the all players send also transmit the user's own position? Yes. > Do players overlap when starting, or do they all start in their own > locations? Either way. > Can players join a game while it is running? Can you join a tournament when > it has already started, etc.. Players can join while it is running, yes. > What does the server send when: the match is over, the tourney is over, and > when you are dead. The socket connection is closed. > If a player dies in the middle of a game but are not killed by falling in > water, what does the "body" count as, or does the space become empty? It disappears from the game (we are clarifying this in the rules now). ----------------------------------------------------------------------- Fri, 30 Aug 2002 15:35:24 > But it does not indicate what the response is if no robots ACTUALLY move. By way > of example, in your example of pushing and bidding that contains a single wall > space (the third example), if neither 1 nor 2 held any packages yet, what would > the server response be in the case where 2 moves first? In that case, neither > robot moves and neither drops or picks up a package, so the server has nothing to > report. Will it report the coordinates of each robot in this case? From the task description: " A robot is listed if and only if it was alive at the beginning of the turn". That is, it is listed even if it didn't do anything. For example, a possible response line is: #1 #2 Which means that there are two robots in the game, nothing happened to them in this turn. ----------------------------------------------------------------------- Fri, 30 Aug 2002 15:49:26 > We have a few questions regarding the rules: Do packages only originate at the > home base? initially packages are at home bases. > Can there be multiple bases? yes. > Will the list of packages available at the robot location always appear as > exactly one line, regardless of length? yes. > Does the server wait for all initial robots to connect before starting the > simulation? no, robots can connect throughout the game. ----------------------------------------------------------------------- Fri, 30 Aug 2002 15:51:38 > If a robot is pushed and drops a package as a result, will the > end-of-turn status update sent by the server include a 'D id' entry? yes > For example, if robot 1 moves west, then is pushed east, dropping > package id 42, will I see: > Scenario 1: "#1 W D 42 E" this is what happens. ----------------------------------------------------------------------- Fri, 30 Aug 2002 15:51:40 > - In the "scoring" section, you give upper bounds for game dimensions, > number of packets, and amounts of money. Could you please give bounds > for the weight of a packet and the carrying capacity of a robot? Bounds for these are 10^9. > - In multi-robot games, can we assume that all robots are given the > same carrying capacity and the same initial amount of money? You can assume all of this, yes. We're clarifying this in the rules. > - Concerning the CPU time limitations at the end of the document, are > these hard limits, and if so could you please specify exactly the > criteria used to determine that a robot runs out of time (and > presumably loses)? E.g. "if after N >= 100 moves a robot has used more > than 1.5 * N seconds of CPU time, it dies", or something like that. It's not a hard limit. After a player has finished a game, the average time/move is calculated, and if it's above the limit, the judges may consider disqualifying the player in that game. > - And just to split hairs: when you talk about sockets (for > communicating with the server), you mean TCP sockets, right? TCP sockets, it is! > Thanks for your time, and congratulations for a really scary-looking > task :-) Thanks! Hope this reply helped... :-) ----------------------------------------------------------------------- Fri, 30 Aug 2002 15:52:13 > Can we always assume syntactically correct input from the server? yes. ----------------------------------------------------------------------- Fri, 30 Aug 2002 15:56:54 > How do you intent to enforce resonable usage of ressources? I've been in a similar > competition before. When people realized their opponent would be running on the > same machine as they were, they started using incredible amount of RAM, just so > their opponent would swap. we will enforce the limits and eliminate misbehaving programs. the limits are not hard though. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:02:37 > What happens when a robot dies? Is it removed from the map? yes > Is the connection closed then? yes ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:02:59 > For every game is there a single home base shared among all robots, or does each > robot have a home base known only to itself? All robots have access to all home bases. There may be several home bases. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:06:24 > How does the robot who gets pushed know which package they have dropped? it is part of the server's reply at the end of the turn. > Do the packages a robot is carrying count as being "in the current location" > and are sent at the beginning of each turn? no, only packages on the ground count as being in the current location. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:07:18 > If a robot dies, does it keep the score accumulated so far in that game, or does > it lose it? Yes. > Also, the task webpage appears to have reverted to the originally posted > version. Hmm. Hopefully a temporary glitch... ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:13: > I'd like know details of the networking part. What calls should one use on > linux? Is there a tutorial of any kind? Or is there another program which > can handle networking? I fear I may use all my time, implementing the > network component, and not the interesting part of problem. there are wrappers to do that kind of stuff. check google. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:17:09 > First off all: lovely task this year. Now for the question: in the description of > the task it is stated that "[t]he game ends either when all packages have been > delivered or all robots are dead." What is want to know is how you can be certain > that this will eventually be the case. because every commands costs at least 1 unit of money, and there is only a limited amount of money. robots die when they run out of money, so eventually everyone will die (like in shakespear's tragedies) ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:09:47 > You say, > "If they are dropped on their destination, the disappear from the game, > otherwise, they stay on the square where they were dropped until a > robot picks it up. See the section The Packages in the task description." > > Can robot A pick the package droped by robot B? Yes. > If yes, will robot A at B'base know robot B's package list, and pick it away? Not sure what you mean, could you give an example. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:17:30 > when the server replies that a robot dropped a > package, is there any distinction between whether the > package was dropped at its destination (and therefore > was removed from the world) or not? There should be enough information in the server replies to determine where packages were dropped, and thus if they were dropped at their destination. > can new robots be introduced into the world at any > time during the game? Yes. > if not, what does the server > reply in the example "#1 X 1 Y 1" mean? It indicates the position of the robot, even if it isn't a new robot. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:17:48 > Are robots necessarily given sequential ID numbers to start a game? No. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:20:26 > > It is a consequence of the rules as posted that games with a single > > robot are pointless: the single player will always win, and only the > > number of wins counts in the overall competition. Is this intended? > > I see that you changed the scoring rules. I note that with the new > rules, it is irrelevant who the winner of a game is; also, there is a > strong disincentive for any fighting between robots. That's your interpretation. :-) ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:21:16 > A small question (or two): when a robot dies, how can its client tell > and what should it do? the connection to the client is closed. > I imagine it could look in the server response > and assume it's died if it can't see itself; is there any other > indication (like the socket being closed)? Should the program exit by > itself, or will it be killed? if you want to be nice to us, you should exit on your own :-) > And I have an ugly BASIC maze-generation program copied from a 1978 > book, resulting in the board at http://swg.ultrastore.com/misc/maze.txt > - not particularly interesting, but possibly a good test for path- > finding... Now I merely need to get a robot actually working ;-) cool!. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:23:11 > Should we use a TCP/IP socket protocoll? Yes. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:24:47 > If a bot dies in a given turn, then will we see all of its packages dropped in > the last server response before it disappears? when a robot dies, its packages are not dropped but instead they disapper. > Also will we see packages that are dropped as the result of a push? yes. ----------------------------------------------------------------------- Fri Aug 30 16:26:45 > > 1) The response format implies that there will be limited visibility, > i.e. a robot will only things happen within a certain distance of it. > > Is that correct? How big will that distance be? Will it be affected by > walls, water, bad weather or just distance? There will be no limited visibility. > 2) Do the N E W S responses sent back reflect what the robot *intended* to > do or what actually happened to him? Here I'm thinking of pushing: a robot > could theorethically be pushed multiple squares in one turn. What actually happened. ----------------------------------------------------------------------- Fri Aug 30 16:30:33 > - What happens if you move to your package destination, then get pushed > (the same turn) so that the package is dropped there. Is it delivered > then? By me or the pushing opponent? Or must someone pick it up and the > immedeately drop it again? It is not delivered, it is dropped on the new square, and can be picked up. ----------------------------------------------------------------------- Fri Aug 30 16:33:11 > Are there telefrags? No. If a new player join on top of an existing player, they temporarily share the same square. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:32:13 > 1. In the server reply for every round, what happens to robots > doing a "no-op" (DROP 0 if not carrying package with id 0)? > Are they listed > as "#9 D 0" ? (assuming robot #9 did the no-op) > as "#9"? > not at all ? (they appear dead?) the second one, i.e. #9 (so robot 9 was still alive, and didn't do anything) > 2. Regarding disconnecting; > Will a robot be disconnected when it is dead > (unable to observe the rest of the game)? yes. > If a robot disconnect, will the game still go on > and the disconnecting robot aren't mentioned at all > in later server replies ? if they disconnect they die. > 3. Just to be 100% clear on one thing, the rules state: > " [...] It is also OK to pick packages which are not there and drop packages which you are not carrying. In both cases nothing will happen." > Will the robot still pick up specified packages which are there, > or will really *nothing* happen? it will still pick up the other things. > 4. Is it allowed to write and maintain files (even between games)? you can create files in the current directory. you are not allowed to use them inbetween games (they will be deleted). > 5. About home bases, can anything be assumed about the initial amount of packages there? you should assume nothing. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:34:28 > The word 'individualized' in the phrase phrase "At the end of each turn the > server will send an individualized reply to each player showing what > happened to all robots during the turn." seems to imply that there is some > sort of line-of-sight, or other factors limiting visibility. Will everyone > see everything? If there are limitations on information, what are they? There is no restriction in sight, we will change that phrase in the description. Thanks for pointing it out. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:35:07 > Just a short question about the client-server protocol: when the > client sends the string "Player" , should a newline be sent at the end ? yes. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:37:03 > If there are no packages available at the player's current position, how is > it supposed to find a place having a package? Will it just search randomly? packages are initially at homebases, so this is a good place to check. also you could keep track of where other robots dropped packages. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:39:53 > 1. Does this limit include statically allocated memory (and program memory) or > just dynamically allocated memory? The limit is on the total amount of memory. > 2. Are programs advised or required not to use more than 64MB of memory? It's not a hard limit, but if the limits are exceeded during a game, the judges may consider disqualifying the player in that game. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:40:06 > At the unique ids for the robots and packages "dense" i.e. allocated from 1...n > with no gaps. no. > What's the maximum number of robots and packages? no limit to the number of robots, the limit for the number of packages is 1,000,000,000. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:41:17 > I have come across some questions you might or might not want to > answer in the FAQ. > > These refer to Document version 1. > > * Can additional robots enter the game at any point in time? Yes. > - If no: What's the point of single-robot games as described in > "Scoring"? The sole robot will always be the one with the highest > final score, so it will win that game. Even if it dies due to a > malformed command or something. > > Also, if no, I think the fourth example in the section "Server's > reply to commands" needs some explanation. Under what > circumstances can a robot appear on some location and move to some > direction in the same round? > > * What are valid characters inside robot identifiers? Are they > integers, perhaps even natural numbers, perhaps even consecutively > starting from 0 (or 1)? A similar question goes for packet IDs > (only they are already specified to be integers). They are integers, no guarantee that they are consecutive. > * If a robot carrying a packet has arrived on the destination square > for that packet but then gets pushed and drops it because of that, > which robot gets the points for the packet? The one who dropped it. > * What's inside the packets? Just curious. :) Great question! We might put it up as a FAQ ;-) ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:42:59 > If a robot dies, is there a way for the server to notify the other > robots? (It's obvious if the robot lands on a death tile, but what if > the robot runs out of money?) I can see strategy being very different > if you know you're the only robot in the game. Perhaps the > "#x X 0 Y 0" message would be appropriate? The number of players can be inferred from the current response from the server. > > If a robot dies, does it disappear? It disappears. > > For example, using the board below, if a robot gets to the home base, > and then exhausts its money (and dies), can another robot move onto > that square to pick up the package? (The dead robot, if it doesn't > disappear, is wedged into the corner, and can't be pushed.) > > 7 5 > ....... > @...... > ##.~~~~ > ...~~~~ > ....... > > Likewise, it would be helpful if the other players could know if a > package was delivered. (ie. is there a message the server could send > to indicate that a package has been removed from the game?) > > Thanks! That would have been helpful, yes! :-) ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:47:17 > 1. If we want to dynamically check our time usage, is real-time a > reasonable approximation for CPU time? i.e., will our player get > more or less the whole CPU? no, robots will run on the same machine so real time is not a good approx. we are using linux so you could use a system call. > 2. When you say "1 CPU second per move on average", is that a mean > CPU usage of 1 second? For instance, would it be okay to > occasionally use (say) 10 seconds in a round, but othertimes use > much less, so that it averages out? it is on avergae, so it is ok to use more time sometimes, as long as you use less time later. > 3. Can we work "in the background" between rounds, and if so does > that just count towards our total CPU usage? your CPU usage is the total for your process and its children. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:48:34 > What exactly mean "one cpu-second per move on average"?? it's a guideline for how long you should plan to spend per turn. > Is "1 drop" the same like "do nothing and spend 1 coin"?? yes. ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:54:04 > I understand that during intialization, the player will be notified of all "home > base"s which contain the packages. this is not the case. > But my doubt came from the fact that, "at every round the server sends the list of > packages at the player's current position". I think the player can calculate that > from "server's reply to commands". Could you please tell if there is any special > reason for this. this is also not the case as the clients cannot tell if the packages dropped by other robots ----------------------------------------------------------------------- Fri, 30 Aug 2002 16:59:19 > If there are 10,000 packages in a game, does that mean they will all show up on > the home spot at the begining, and thus that we can expect up to 40,000 integers > to come on a single line (packages available)? yes :-) > Sorry if this is the wrong place to send such questions; > I haven't bumped into any obvious pointers for where they > should be directed.) this is the correct place, but please also check the FAQ and the webpage before posting to us. ----------------------------------------------------------------------- Fri, 30 Aug 2002 17:07:42 > How many robots maximum will participate in a game? there is no limit to that ----------------------------------------------------------------------- Fri, 30 Aug 2002 17:24:27 > * What hard time-outs exist in the protocol? What does "Aim for" > mean? What does "average" mean? It's a soft deadline. If after a game, it turns out that a player that made N turns has used more than N cpu seconds, the judges may disqualify the player in that game. > * If a robot is pushed more than once in a turn, does it drop a > package for each push? Yes. > * How many robot clients can one team submit? Only one (this will be clarified on the web pages). > > - Will they ever play each-other? > (N/A) > * Is a step into a wall malformed, a no-op, or something else? A no-op. > > * There appears to be no way to be certain of another robot's current > amount of money. Is this intended? Yes. > > * When a robot moves into water, does it occupy that space? > > - For the rest of that turn? > > - On subsequent turns? > > - Not at all? It dies, and is removed from the board. > > * Is it likely that robots will join a game that is already well > advanced, or will robots all be loaded at the start of the game? Robots may enter at any point in the game. > * You say, "The winner of the competition is the robot with the > highest total score over all games." If a robot dies, can its > client join successive games? Thus, is the winner the *client* with > the highest total score over all games? Or is the client > disqualified if its robot loses a single game? In fact, how could a > *robot* play in multiple games? There is nothing in the protocol to > tell a client that its robot is now in a new game. That should have been the *player* that has earned the highest total score, not the robot. We will change that in the description. So a robot that dies does not get connected to a new game. > * If another robot dies during turn i-1, its death may not be evident > in the end-of-turn summary. (Perhaps it has run out of money.) > Instead, its death may only become evident when it fails to appear > in the end-of-turn summary for turn i. Is this a correct reading? Correct! > * The examples for end-of-turn summaries show a robot joining and > moving in one turn. Suppose such a summary was for turn i. Three > questions: > > - Does this contradict the statement that a robot will appear in > the summary "if an only if it was alive at the beginning of the > turn"? A joining player is by this definition alive at the beginning of the turn, so it is not a contradiction. > - From the point of view of robot 1, did it: > * receive the end-of-turn summary for turn i-1 (in which it > was not present); No. It only gets the location of other players as part of the initialization protocol. > * then bid with the knowledge of existence of all the other > robots; > * then see its version of the end-of-turn summary for turn i? Yes. > - Does this mean my (already in the game) robot 2 must have bid on > turn i without knowing it was bidding against this other > (newly joining) robot 1? Yes. > * What is the range for the maximum carrying capacity? At most 1,000,000,000 > * What is the range for the weight of 1 package? The range is 1 to 1,000,000,000 > * Is the client program freshly installed between each game? Yes. ----------------------------------------------------------------------- Fri, 30 Aug 2002 17:29:52 > Can new packages appear during the game or are they all > fixed at game start? packages will not appear during the game. > In particular, if a robot has seen that a home base is empty, can it assume that > it will remain empty (except robots dropping packages there)? this however is not the case - a robot could go and drop some packages on a homebase and then it won't be empty anymore. i am not sure why they might want to that though. ----------------------------------------------------------------------- Fri, 30 Aug 2002 17:34:16 > Re the test server: can we have an estimated time at which the server will be > available. We hope to have it up and running within a couple of hours. > Can we use it to play variants of our program against each other? Yes. We will set up some single-player game servers and some multiple-player game servers. > (i.e. do we each get to run our own instance of the server?) We will provide a binary for download, so you can run private games. > 1. What is the range of legal package weights. > Presumably >= 1? What is the max? We will not use weights exceeding 1000000000. ----------------------------------------------------------------------- Fri, 30 Aug 2002 18:09:32 > In considering strategies for this task, I've hit a snag. Perhaps I'm not interpreting the rules correctly. Please consider the following situation: > Initial board (North is up): > ...... > ..@12. > ...... > Move 1: both robots move West > ...... > ..12.. > ...... > (Robot 1 is on top of 1 package) > Now, robot 2's best strategy is to push robot 1. Robot 2 can make this move with bid=0, because even if robot 1 goes first and executes a pickup, robot 2 will push robot 1 causing her to drop the package. > Move 2: Robot 2 pushes West > ...... > .12... > ...... > (Robot 2 is on top of 1 package) > Now, the above situation is repeated for robot 1. Her best strategy is to push robot 2. It seems that when this situation occurs, there will be one of two outcomes: > 1. infinite loop > 2. one of the robots will have to concede the package to the other robot. > (If there is a limit on the number of turns, this feels like a Prisoner's dilemma!) > Have I misread the rules? you are correct that this kind of thing can happen on some maps. robots may not bid 0 ----------------------------------------------------------------------- Fri, 30 Aug 2002 18:10:25 > if a robot is pushed and drops a package how will it be reported? the same way other dropped packages are reported. -----------------------------------------------------------------------