Automation and Scripting - An investigation into further abuses of BGS and Powerplay

Status
Thread Closed: Not open for further replies.
No offense but you are starting to be as condescendant than the one putting willingly bot in open.

How was my comment in any way condescending? I answered why we're discussing a bot to dock. I put forward a conjecture, and a hypothetical. The rest is just an academic exercise. Condescending?
 

sollisb

Banned
How was my comment in any way condescending? I answered why we're discussing a bot to dock. I put forward a conjecture, and a hypothetical. The rest is just an academic exercise. Condescending?

If you think removing the docking computer will make all the botters leave, then frankly, you are deluded.
 
If you think removing the docking computer will make all the botters leave, then frankly, you are deluded.

That's rich from someone who thought all they had to do was centre the orientation radar and they were practically docked.

I'm sorry you feel so bitter, man, but I did PM you to keep any disaffection away from public eyes. When you keep making ridiculous comments and spread misinformation on this thread, you should expect people will question you. So when you said "Lining up with the station is easy peasy" and I used PMs to show you things you hadn't realised, I don't see why you're so bitter about it.

Which brings me to the next point which after re-reading I now realise must be addressed. You said on this thread...

OMG, you're being such a jerk!
You were asking for specifics on how it could be done. FDev you have my explicit permission to read my PMs to Merkir. In those messages I explained how easy it was. he kept asking for specifics, and I grew ever concerned that he actually was trying to create a bot himself. He also threatened to post on the forums that I was full of manure if I didn't explain to him. So I ended the conversation.

These are bald-face lies.

I would like you to either retract this, or publish our PMs in their entirety and highlight the parts relevant to the above, or give me permission to post them. Please go ahead.

In particular I'd like to know why you say "He also threatened to post on the forums that I was full of manure if I didn't explain to him." I can assure you I have no interest whatsoever in your bot ideas, but I do care about misinformation.

Was it due to the following PM after you had realised your easy solution wouldn't work?
Merkir said:
So you've posted "Firstly; Lining up with the station is easy peasy."

Since you seem to have no idea how it's easy, are you going to post a correction? I hope you have the common decency to do that because leaving misinformation is not cool. If you don't, I will.

And since you didn't, I thought I went very easy on you by simply posting here that you had a misunderstanding of the HUD orientation radar.
 
Let's stop there. I'm not sure how reliably a bot could "turn towards the position 10 km in front of the slot". As a human, easy, but how does the bot know the concept of "front of the slot"? So to answer that we can look at the station hologram orientation. The issue I see here is that pattern recognition from an unknown 360 degrees orientation is flaky at best from off-the-shelf script bot software.
Probably the easiest approach to automate from on-screen data, I think, would be

1) Exit supercruise
2) Target the planet, orient to it, fly for about 15km.
3) Re-target the station. The ship is now guaranteed to be roughly in the front 60-degree cone of the station, possibly much closer.
4) Make orientation changes while stationary to line the station hologram up with a pre-compiled reference. This is probably the hardest bit to get right, but on an Orbis/Ocellus a semi-random walk strategy might do reasonably well.
5) Use roll and lat/vert thrusters, together with the compass, to move the ship onto the station's primary axis.
6) Move forward, request docking clearance, match rotation and go through the slot.

The big catch with this: you're spending a lot of time in the normal space instance of the station. These things tend to get busier with NPC traffic over time, so the chance of a ram costing you the ship - or even just forcing a log-and-restart - goes way up.

Given just how long it took for Frontier to get the docking computer to work reliably - with full access to position, orientation and velocity data for everything in the instance, as well as the ability to write to the docking/undocking queues so only one ship tries to traverse the slot at once - I expect it would be extremely difficult to get right quickly and reliably.


At any rate, Frontier aren't going to get rid of the Docking Computer. Increasing the frequency and hostility of interdictions would be a more reliable method but again not very popular...
 
So, just to make these two have a break.. our ship can autodock on stations and on planets (don't forget this), but there is no autodepart on any of them.
Oh wait, when you dismiss the ship, it leaves the planet in auto pilot, and NPCs can dock in almost any pad of any of the stations, but also leave them without any issue.

We are getting there.. there is a deliberate reason why the players don't have auto-pilot to leave stations or to land/leave planets from FD, and honestly, I'd leave all these as they are.

Lets assume coding docking and departure is not hard. Same as piloting in SC, not hard. Now, when it comes to go from A to B, being A and B unknown.. a bot has several extra problems to deal with, different star sizes, different routes (with and without refuel), jump limits because of weight, going through different states, different security levels, etc.

Despite docking and departing, or even flying in SC being easy, not knowing what the route looks like adds the uncertainty to prevent bots from flying 24h, or limit them to 'known activities' that can be tracked sooner or later.
 
Two (technical) questions then:-

1) When one of these bots was taking off from the pad. If you got in the way so their upward thrust fail at some point, am I right in saying they'd then fail to even make it out of the station? ie: They're as dumb as thrust up for X seconds... Thrust forwards for X seconds?

2) How on earth are they doing all the SC'ing and indeed even when leaving the station lining up for the hyperspace jump? Are these scripts using screen scraping and optical recognition and the look to understand the HUD/interface?
 
1) When one of these bots was taking off from the pad. If you got in the way so their upward thrust fail at some point, am I right in saying they'd then fail to even make it out of the station? ie: They're as dumb as thrust up for X seconds... Thrust forwards for X seconds?
Probably, but it might still try to realign if the screen recognition doesn't see a clear slot.
2) How on earth are they doing all the SC'ing and indeed even when leaving the station lining up for the hyperspace jump? Are these scripts using screen scraping and optical recognition and the look to understand the HUD/interface?
yep, the hollow means wrong direction, the white dot in the center means good, other than that: pitch and yaw.
 
Are these scripts using screen scraping and optical recognition and the look to understand the HUD/interface?

Almost definitely, yes. This kind of optical recognition is fairly straightforward — one would only need to use the compass to get the right heading.
 
Probably the easiest approach to automate from on-screen data, I think, would be

1) Exit supercruise
2) Target the planet, orient to it, fly for about 15km.
3) Re-target the station. The ship is now guaranteed to be roughly in the front 60-degree cone of the station, possibly much closer.
4) Make orientation changes while stationary to line the station hologram up with a pre-compiled reference. This is probably the hardest bit to get right, but on an Orbis/Ocellus a semi-random walk strategy might do reasonably well.
5) Use roll and lat/vert thrusters, together with the compass, to move the ship onto the station's primary axis.
6) Move forward, request docking clearance, match rotation and go through the slot.

The big catch with this: you're spending a lot of time in the normal space instance of the station. These things tend to get busier with NPC traffic over time, so the chance of a ram costing you the ship - or even just forcing a log-and-restart - goes way up.

Given just how long it took for Frontier to get the docking computer to work reliably - with full access to position, orientation and velocity data for everything in the instance, as well as the ability to write to the docking/undocking queues so only one ship tries to traverse the slot at once - I expect it would be extremely difficult to get right quickly and reliably.


At any rate, Frontier aren't going to get rid of the Docking Computer. Increasing the frequency and hostility of interdictions would be a more reliable method but again not very popular...

Nice solution. It's good to see a variety of ways to go about it. I'd forgotten that most stations are oriented about 45 degrees to the planet. Again it probably comes down to whether the software can handle pattern matching of highly variable positions, where rotation and scale is unknown. I'm not conversant with anything other than two products and both only match with no allowed change of angle or scale, so as you say, some kind of random walk would be required. Does anyone know or use more sophisticated pattern recognition software?
 
Almost definitely, yes. This kind of optical recognition is fairly straightforward — one would only need to use the compass to get the right heading.

And if planets are in the way when SC'ing to their target? Clever stuff!
 
Two (technical) questions then:-

1) When one of these bots was taking off from the pad. If you got in the way so their upward thrust fail at some point, am I right in saying they'd then fail to even make it out of the station? ie: They're as dumb as thrust up for X seconds... Thrust forwards for X seconds?

No. As I noted before, you can clearly see the Cutter in the video stop to let a NPC pass. So there is at least a degree of obstacle recognition.
 
Again it probably comes down to whether the software can handle pattern matching of highly variable positions, where rotation and scale is unknown.

I'm not an expert when it comes to recognition software, but given the state of the art in terms of autonomous cars, face recognition and other things I'm really confident there's a software out there that is able to calculate the position of this little blue dot/circle:

h1uFQ1N.png


You can even filter out the reddish stuff and there'll only be your direction dot left. You know the center of the globe, you know the position of the dot, you know the radius of the globe, you know whether it's a filled dot or a hollow circle. What else do you need to calculate the direction you want to head to?

If you get a reliable algorithm for the globe thingy and take target distance and speed into account (so you know when to speed up, throttle down, or when there's something unexpected happening like getting stuck in the slot because a T10 suddenly showed up, forcing you to emergency-log), then you should be good for 90% of the whole journey. The other stuff may be special cases that need more detailed algorithms or even player interaction. But you need a computer to run the game instance and the bot anyways. There's always the possibility of keeping the bot out of the game until a player can check and solve the situation manually.

Given the fact that there are maybe a couple dozens of bots and not a couple hundred, I'd say the abusers know they have to check the bots constantly and be able to solve situations. The scripts are obviously not perfect. So there's no need to discuss any possible issue here - there's already a lot of damage just by covering 90% of what humans usually have to do manually.
 
You can even filter out the reddish stuff and there'll only be your direction dot left. You know the center of the globe, you know the position of the dot, you know the radius of the globe, you know whether it's a filled dot or a hollow circle. What else do you need to calculate the direction you want to head to?

I wonder if the bots are using certain hud colours to make it easier for visual recognition software to pick up the hud cues.
 
I must say. As a hillbilly redneck that uses duct tape for everything. Thats a whole lot of trouble to go through to cheat in a video game.

Actually it takes literally a minute or two to open up the graphics config file and change the HUD colors to something with high contrast from the rest of the game.
 
I'm not an expert when it comes to recognition software, but given the state of the art in terms of autonomous cars, face recognition and other things I'm really confident there's a software out there that is able to calculate the position of this little blue dot/circle:

https://i.imgur.com/h1uFQ1N.png

You can even filter out the reddish stuff and there'll only be your direction dot left. You know the center of the globe, you know the position of the dot, you know the radius of the globe, you know whether it's a filled dot or a hollow circle. What else do you need to calculate the direction you want to head to?

If you get a reliable algorithm for the globe thingy and take target distance and speed into account (so you know when to speed up, throttle down, or when there's something unexpected happening like getting stuck in the slot because a T10 suddenly showed up, forcing you to emergency-log), then you should be good for 90% of the whole journey. The other stuff may be special cases that need more detailed algorithms or even player interaction. But you need a computer to run the game instance and the bot anyways. There's always the possibility of keeping the bot out of the game until a player can check and solve the situation manually.

Given the fact that there are maybe a couple dozens of bots and not a couple hundred, I'd say the abusers know they have to check the bots constantly and be able to solve situations. The scripts are obviously not perfect. So there's no need to discuss any possible issue here - there's already a lot of damage just by covering 90% of what humans usually have to do manually.

You are pretty close with your analysis regarding the ship's compass. You don't even need to know if the blob is hollow or not. If you pitch up and the blob moves in the expected direction then the blob is indicating a target in front of the ship otherwise the reverse.
 

raeat

Banned
So, it has been established botting is happening.
We have plausible reasons for why, and there is certainly precedent for humans going to great lengths to cheat in competitive games.
And the last few pages seem to have been unravelling some of the possible hows.
We've even had postings indicating sites that tell us where we can get the scripting help.

Does anyone still have any doubts that this botting represents a serious problem?

So, the only question that remains is, "How serious a problem?"
I contend that this sort of botting will end meaningful human involvement in the BGS and Power Play. Some think that is somewhat hyperbolic, but if a bot can move many times as much goods in BGS, and papers in PP as will make player player actions irrelevant, then I think there is quite a reasonable argument to reach the conclusion I have. If these game modes are to be at all meaningful for human players, then the bots must be dealt with.

Now, many players don't care about the BGS or Power Play. Fair enough. It is your choice what you want to participate in in the game, but there are players who want to participate in BGS and/or Power Play. Clearly, FDev intended these game functions to be available for the players (at least Power Play - arguably, for BGS).

So either FDev makes bots available for everyone to use or for no one. The cats out; there is no going back now.

So, FDev. What are you going to do about this rising problem? Whatever it is, best act soon ...
 
Last edited:
Status
Thread Closed: Not open for further replies.
Back
Top Bottom