There are plenty of great programmers about, unless the pay is competitive (because these guys can name their own price) they won't take it up. Plus, other countries like Canada have highly aggressive sweeteners for these skilled jobs to go to Canadian dev houses.
AI Vehicle Programmer .... the search continues
- Thread starter Lightspeed
- Start date
There are plenty of great programmers about, unless the pay is competitive (because these guys can name their own price) they won't take it up. Plus, other countries like Canada have highly aggressive sweeteners for these skilled jobs to go to Canadian dev houses.
I agree but would it be a bad thing if the NPC drivers weren't very good and crashed a lot? Might generate some gameplay in getting mats and cargo from these crashes. And lets face it, it would be quite funny watching the NPCs drive off the odd cliff or get caught between to outcrops or whatever
Playing old games like Battlezone (or the new versions) shows this is fun for a while, but breaks immersion when a tank wants to bury itself in a cliff-face. But if the AI is done right you can make then as dumb or as intelligent as you like.
Which reminds me: FD have also got to sort out ship the landing gear floating above the planet surface.
I agree but would it be a bad thing if the NPC drivers weren't very good and crashed a lot? Might generate some gameplay in getting mats and cargo from these crashes. And lets face it, it would be quite funny watching the NPCs drive off the odd cliff or get caught between to outcrops or whatever
Would be fun - for about five minutes.
No, IMO a benchmark test for the AI driver would be the Pomeche Ridge Challenge.
Quite sad about this (that the placement hasn't been filled yet) ... I had high hopes that this coder role in combination with the "base building" leak suggested that planetary bases were going to become much more alive (with NPC ground vehicles going about their business a bit like ships around a starport) and that we might end up with surface based missions and objectives. If they haven't even found a coder for this yet tho those dreams seem somehow less likely for 2020. Perhaps there's still time?
LOL - yeah, any NPC SRV that could post even a mediocre Pomeche Ridge time would be impressive ... most impressive. Perhaps FD should rephrase the ad? Make it like a competition ... can you code a vehicle AI to complete this challenge? Winner gets a job at Frontier!
np, why? would you accept any job you're qualified for? there's lots of other potential reasons for candidates to back down, e.g.:
- the salary is just crap and some people just won't work for 'benefits' like free fruit and coffee
- don't want to get involved with another crappy inhouse 'genius' game engine and prefer to stick to industry standards
- they just distrust hipster offers or have detected unreasonable 'i have no clue what i want but i want it done' culture
- they smell other types of toxic relations (white elephants, entitled gurus, disruptive management, etc).
- they simply get undermined by someone opposing that application from the inside because reasons, usually political
saying that 'ai driver programming' is a rare cutting edge skill in 2020 is just ridiculous. if they haven't covered this position it is most likely not because lack of professionals but because something is wrong on their end or on what they are offering, or asking for.
Pathfinding on/ in predefined areas is easy. Its the crazy possibilities of planet surfaces thats the biggest problem, because they are built to rules that are not AI friendly. As an example: how far ahead would an AI SRV need to 'look'? The AI would need to understand the whole planet in advance to know what routes are traversable, and not a dead end canyon after traveling 20Km along it.
This might illustrate how intense it can get:
And this is in predefined areas.
but then we would need another class 1 limpet controller
With huge wages no doubt.
No, just fit the decontamination controller and put it on the road.but then we would need another class 1 limpet controller
but then we would need another class 1 limpet controller
could you push it around with a class 1 abrasion blaster?
i dunno maybe Class one small seismic charges may get it moving?
I'd consider that "cheating". The AI should have (and work with) the same terrain information as the player (not necessarily in the same form - as all the data is already there, there would be no need to first map the terrain to image data and then run a vision AI over the display image). Even then, it cold be faster and make a more comprehensive use of the available data (the SRV radar is actually 3D) - similar to the ship AIs currently having a much more refined thruster control than a human pilot would usually manage.
Yes, some players team up so you can have a scout in a ship to select the coarse route from above, and the diver in the SRV trying to follow that route - but then again, there's no reason why a couple of AIs couldn't fulfill the same roles, with similar limitations. But otherwise, human drivers also occasionally end up in a dead end. So you either have to find a way around or over, or you need to track back. That part - "find a way around or over" - should be the real fun, as it involves (sometimes) creative solutions, which are a real challenge to any AI so far.
Ship AI is (relatively) easier because space generally has less obstacles- EDs AI struggles with asteroids / debris / gravity though as I suspect its not really set up for it. If it was bots for CQC would be a thing. You also have to make the AI 'cheap' so that it does not hog resources.
Its an interesting problem for sure. But I do know that retrofitting is harder to do than adding an planned extra.
S'funny really...
I wouldn't have thought AI vehicle programming would be a particularly specialised thing.
I start off thinking of all the off-road driving games which require it but then it occurs to me that even things like RTS's require AI vehicle programming for path-finding etc.
Thing is, the vast majority of games take place on pre-rendered terrain, so an AI vehicle can just look at the data for the map to figure out how to get around.
In ED, with it's procedural planets, an AI vehicle would need to be able to analyse it's surroundings and figure out the best way to get around on the fly.
I guess that probably makes the task MUCH harder than it is in the vast majority of other games?
Anybody know of a driving sim or RTS that does take place on procedural terrain?
I wouldn't have thought AI vehicle programming would be a particularly specialised thing.
I start off thinking of all the off-road driving games which require it but then it occurs to me that even things like RTS's require AI vehicle programming for path-finding etc.
Thing is, the vast majority of games take place on pre-rendered terrain, so an AI vehicle can just look at the data for the map to figure out how to get around.
In ED, with it's procedural planets, an AI vehicle would need to be able to analyse it's surroundings and figure out the best way to get around on the fly.
I guess that probably makes the task MUCH harder than it is in the vast majority of other games?
Anybody know of a driving sim or RTS that does take place on procedural terrain?
It would certainly be a fun problem to have a go at from an AI standpoint.
Stressful as heck on a deadline and with much hanging off it.. I wonder what their fallback solution to this is?
Last edited:
Hmm. The requirements are tough.
During my universtiy time in the course of neuronal networks, i was one of several teams competing in writing control software steering a car through a track. But the tracks were in the Doom engine, so effectively only a 2D environment. Getting that done was tricky enough. And that's even just one part of several requirements, which are not easy to meet. And if the technical requirements were not bad enough, it also is in England.
The funny part would be there only for a short while. Then we'd just be stuck with yet another part of the game, which doesn't work properly. Especially note player mentality: i remember when the first videos of the actual Mechwarrior 5 were released. Everything still in the alpha, with clearly early stage vehicle AI.
So what were people talking about? How great the updated mech models looked? How awesome all the destructive terrain was? If they liked the HUD or did not like the HUD, compared to MWO? If the weapon effects on the new engine was good or not?
Not really. All the discussion was about how terribly the vehicles moved, crashed and got stuck all of the time, usually followed by declaring the developers to be incompetent. And that, as mentioned, was for videos of an early alpha of that game. The bar is much higher even for delivered software. Bad drivers would not be good for the game.
During my universtiy time in the course of neuronal networks, i was one of several teams competing in writing control software steering a car through a track. But the tracks were in the Doom engine, so effectively only a 2D environment. Getting that done was tricky enough. And that's even just one part of several requirements, which are not easy to meet. And if the technical requirements were not bad enough, it also is in England.
I agree but would it be a bad thing if the NPC drivers weren't very good and crashed a lot? Might generate some gameplay in getting mats and cargo from these crashes. And lets face it, it would be quite funny watching the NPCs drive off the odd cliff or get caught between to outcrops or whatever
The funny part would be there only for a short while. Then we'd just be stuck with yet another part of the game, which doesn't work properly. Especially note player mentality: i remember when the first videos of the actual Mechwarrior 5 were released. Everything still in the alpha, with clearly early stage vehicle AI.
So what were people talking about? How great the updated mech models looked? How awesome all the destructive terrain was? If they liked the HUD or did not like the HUD, compared to MWO? If the weapon effects on the new engine was good or not?
Not really. All the discussion was about how terribly the vehicles moved, crashed and got stuck all of the time, usually followed by declaring the developers to be incompetent. And that, as mentioned, was for videos of an early alpha of that game. The bar is much higher even for delivered software. Bad drivers would not be good for the game.
Do you think the delay was in part because they had to shift someone off FCs/bug fixes to do the AI work they are trying to recruit for?
Probably not. I mean there is always that possibility but I suspect the maintainance team is busy enough as is with everything else to then take resources away from them - they are also probably more generalist where as this is a much more specific role with more specialized requirements.