Procedural map generation in Spy DNA

by Jason Sams

It’s been a while since I wrote my last update. I’ve been hard at work on a few things. But as promised in the last update, today we will talk about mission maps.

We have the core of the map generation up and running. We have tested it generating maps from 128 meters square to 2 kilometers.  Map generation times are pretty good; just a few seconds in most cases.  

Procedural map of a wooded rural area with roads and trees

The maps are complete with bushes and trees. We are planning to add grass too, but that is a little harder to do without hurting performance, so it may not make our first Early Access release.

The size of the map will have a large effect on the time a mission takes to complete, and the general flow of a mission. For example, on a small 256 or 128 square meter map, there is no reason to bring any sniper weapons with you on a mission. Most of the regular rifles are “good enough” at those ranges while being much more useful up close.

Closer view of a procedural map, showing transition from sandy to grassy terrain

We understand our players will have a variety of play styles. So we will be adding an option to the settings to adjust the map size to larger or smaller to mirror what you enjoy most. This will apply a +1 or -1 to the map size settings. The supported map sizes are 128, 256, 512, 1024, and 2048 meters. At the default settings all missions will be on maps from 256 to 1024 meters. So applying the +1 would change that to 512 to 2048. 

When you start a mission, you will be able to see all of the terrain. We assume that in the future you'll still have satellites and drones to recon the area before you deploy. Hidden or movable items such as enemy patrols will be hidden by the fog of war until a team member manages to spot them. Once spotted, they will be marked on the map. If you lose contact, the marker will remain at the last position a team member saw them.

With the upcoming demo we will be using the procedural maps to allow the player to generate skirmishes. We want everyone to have a chance to try out our unique combat system and get a feel for the game.

Lumberyard

It’s been a really busy two months here at Shy Snake. We have made a few adjustments to the development of Spy DNA which we think will make for a better game. The big change is we have moved from Unreal to Lumberyard game engine. This also created an opportunity to make a few other smaller changes to the game.

We decided to move to Lumberyard because it allows us to make much more dynamic maps. Unlike other engines it doesn’t require us to pre-bake lighting to get good quality lighting on the maps. This frees us up to procedurally generate maps, allowing for a greater replay value and much more variety in missions. It also allows us to change the maps on the fly in response to player actions. This means a more destructible world. 

We are mostly complete with the port of the game code to Lumberyard. AI is the last major piece left to do. At this point the UI is mostly complete, the game logic code is ported, animations are working, and we integrated the articy code for the dialog system. We are still on the same schedule as previously announced with a public demo / early access, possibly as early as December, but for sure in Q1 2017.

Animation is another area that was impacted by the change. We are switching from in-place animation to root-motion animation. What this means for the players is the characters should move in much more realistic ways. The characters’ feet should slip far less. Also we added start, stop, and turn animations so the beginning of movements will look a lot better. 

The transition required us to re-import all of our art assets. We took this as an opportunity to replace our character models with higher quality models. To do this we switched character creation tools and are now using iClone Character Creator. 

Same character created in Autodesk Character Generator on the left and iClone Character Creator on the right.

Now that we have worked with Lumberyard for almost two months, we have learned a lot about it. If fits very well with our code base. As we have settled in the level of productivity on our game, code is settling at a higher level that we achieved with Unreal. This means we will be able to put more effort into building a good AI for you, the player. 

Now it’s not been all positive. On the art side things are going a bit more slowly. Lumberyard is an editor in transition which has created some overhead for us in making maps. In part this is offset by switching many maps to procedural generation. However, it will be a bit before we can show you the before / after of the player base. Yes, even though we can now procedurally generate maps, we can also still hand-generate them. 

We will have much more on this going forward. I believe the engine transition will allow us to ship a game which is more in line with what our players expect. It allows us to generate a much larger variety of maps, allows you to have a larger impact on the map in terms of being able to destroy objects, and allows us to invest more heavily in AI making the gameplay more interesting.

Demo Complete!

We finished putting the final touches on the demo and are sending it out to a few testers tonight.  Once they have had a chance to play with it and we address any issues, we will send it to a wider list.

We have also decided to make the demo available to anyone that has backed our Kickstarter campaign at Beta or higher.  I.e. those that wanted an early look and were willing to put up with the limitations of early builds.  It’s around 1.4Gb for the two demo levels.  

One of the good side effects of working on the demo is it forced us to address some long standing bugs that we had been putting off.  

Examples:

• Write a manual
• Add objectives to the maps
• Sort the dialog answers so “goodbye” was always at the bottom
• Put the status indicator over the character portraits
• Assign skills to the NPCs

We also added the end of mission screen which includes the status of the objectives, the party, and some statistics about the equipment used.

We also added an “event cam” to the game.  It’s purpose is to highlight important events to make sure the player is fully aware of what is happening in the game.  We will add settings for the final product so players can customize which events are highlighted.  

Changes of note:

• Added check marks to objective list

• Made movement modes available to AI waypaths

• Maps can now specify if the characters start with a weapon equipped

• Hide shooter weapon in gunsight cam to avoid possibility that weapon would block view

• Fixed bullet trails and sounds for some weapons

• Enabled event cam for kill shots

• Fixed AI bug that cause NPCs to sometimes empty the clip into a already down character

• Taught AI to reload

• Added medium and heavy combat armor

• Fixed icons for armor

• Tweaked ROF for a few rifles

• Added new pistol models

• Added descriptions to armor

• Fixed bugs applying armor to hits

• Added keyboard shortcut for previous and next target visible to selected character

• Replaced generic sounds with per-character sounds

• Removed occlusion check for muzzle flashes.  Was causing flicker.

• Greyed character portrait when stunned, unconscious, or dead

• Set skills for all NPCs on demo maps

• Replaced placeholder NPC models with correct models

• Fixed bug causing turns for unconscious characters.

• Reviewed max-texture sizes for small objects

• Added post-mission summary and stats screen

• Fixed bug with default aim point after a melee attack

• Updated 3D queue overlay to fix bug with stuck targeting line

• Fixed gun empty sound

• Hide contact counts when character is unconscious or dead

• Added sorting support to dialog answers

• Changed default to show long answers rather than short summary

• Fixed cursor hitbox for hud to resolve small dead-space area

• Added skill check for spotting and rate limit spot checks

Demo development update

This is Jason with a long[ish] update on our progress towards demos. We have been hard at work getting a demo ready. I wanted to give everyone a feel for what it’s like to develop a game with a small team.

AI 

Shipping a demo with combat requires the AI to behave in a reasonable way. Up to this point they have been heavily scripted, but to make a more interactive demo we have accelerated implementing a few things related to perception and reaction to threats. 

• We now simulate “a quick glance”. If a character hears a noise and the source of the noise is in their line of sight but not field of view, they now have a chance to also visually detect and evaluate the source. 
• We revised “contact tracking”. Previously the fog of war system was too aggressive hiding characters within the line of sight but out of field of view. Now you can track due to sound and neutral and friendly contacts are retained for longer. 
• NPC AI now has separate tactics and actions. This allows them to return to their previous actions when interrupted and make better choices about when to take an interrupt vs continue what they were doing. 
• AI commands can now be issued from the dialog scripting language. This was a temporary regression from switching to the new dialog editing software. 
• Changed vision markers to appear if any party member has contact with enemy, not just selected character. 

Dialog 

Alex has been entering all the dialog for the demo and initial levels. This has exposed a few issues during testing and we have been fixing those as they pop up. 

• Scripts can now be attached to any dialog node. Previously they only worked on player answers. 
• Added support for N-way conversations. Previously dialogs could only be between the player and one NPC. 
• Some sequences would cause the wrong portrait to be displayed in the chat. This has been fixed. 
• Added scripting command to make a NPC face an object. 
• Added scripting command to play a specific animation. 
• Added item list to instruction nodes to allow item transfers between NPCs and player inventory. 
• Characters now play a conversation animation while talking. 
• [todo] Add sorting option to answers Fixed bug where NPC could interrupt an existing chat to start a new one. 

Objectives 

The game has always had a plan to present the player with specific mission objectives. We decided to pull this in [schedule wise] for the demo. So about three days ago I started implementing the objective screen. 

• Implemented an Objectives tab in the character screen. 
• Added a place to edit the objective list in the UE4 level editor. 
• Linked the text descriptions to the description objects in the dialog editor. 
• Added C++ code to track progress of objectives. 
• Added C++ code to display objective state changes (complete or failed) to player log. 
• Added objectives to test map and played though. 
• Fixed bugs resulting from above additions. 
• [todo] UI cleanup of objectives screen. Display check or X for pass and fail. 

Maps 

Alex has also been cleaning up the base map for the demo. We should have done this one earlier because the improvement in graphical quality have been huge. 

• Review post-processing settings for map. 
• Reviewed lighting on map. Added descriptions to items, which will appear in-game as tool tips if you hover over an item. 
• Fix some material issues where the materials were too metallic (too reflective). Floors are never clean enough to behave like a perfect mirror. 
• Fixed window transparency to work with the fog of war. 
• Added AI paths and tasks to NPCs on base map. 
• Added lots of environmental items to make the map look more like a real office environment. 

Misc 

• Fixed bug where weapons could fire in a wrong mode if the player used the default mode. 
• Added contact counts to character portraits, showing how many friendly, neutral, and enemy NPCs a character perceives. 
• Fixed bug with NPC movement that would sometimes use the wrong movement speed (e.g. walk instead of run). 
• Fixed bug where movement animation would not stop when a character finished following another character. 
• Moved item descriptions from in-game editor to external tool to support localization. And yes, we hope to have the resources to localize the game. 
• Fixed character names getting corrupted in event log. 

I hope this gives everyone an idea what two weeks of development here at Shy Snake look like. We operate at a higher than normal velocity and want our backers to know about the progress we’ve made. 

Kickstarter update 4: Stealth

Update from Jason

It’s been a busy week here at Shy Snake.  We just pushed the update to move Spy DNA to Unreal 4.12.  We were waiting until our demo at AFK was complete so as to not break anything right before it was time to show.

I’ve been 100% focused on AI.  One big part of AI is making the system respect stealth.  In Spy DNA we have a system that gives each character attributes for senses.  These, combined with the characters skills, will determine how likely a character is to detect a player.  

Sound:

Everything a player does can generate a noise.  Some things such as sneaking are quiet, while firing a gun would be loud.  The sounds will dissipate over distance and with obstacles.  The sound strength when it reaches the character is used to make a check against the character's ability to determine if they hear it.  If they do then they may react.  For the players, when your character hears a sound, we add a symbol with an arrow to indicate the direction of the sound. 

Sight:

Sight in Spy DNA serves two purposes.  First is to detect a character.  The second is to evaluate the character.  Being a spy will often place you in locations where you will not be overtly carrying weapons.  This means that in an area mixed with friendlies, enemies, and civilians, an enemy would both have to see you and then decide you are a threat.  This means checks against their ability to spot hidden weapons, provided you don’t have a rocket launcher on your back.  Anything in your inventory that could blow your cover is a risk for detection even if it’s not equipped.  Obviously not having something equipped does make it harder for them to detect it.  This creates an incentive to carry light and concealable weapons on many missions. 

The sight AI includes a cone of vision so the enemies have to look in the correct direction to see you. If they hear a noise they will turn to look, provided they're not doing something more important already. The AI also ranks everything it sees in terms of “potential threat.”  A civilian going about their normal business would register near zero.  One running and screaming would start to move the needle.

Smell:

This isn’t really used to detect enemies, but rather hazards.  Gas leak, smoke, or a specific perfume could all be clues or hazards.  

---

Alex has also been busy.  While I have been implementing stealth, she has been working to de-stealth Spy DNA and make sure the fans of thoughtful tactical games know about us.  If you know anyone that would be interested in Spy DNA, please give her a hand in getting the word out.

Kickstarter update 2: Combat visualization

Folks, we're currently on Kickstarter, and sharing a lot of our updates there. We'll be also sharing the updates here for your convenience. If you haven't yet, help us by making a pledge, sharing our project on social media, or be daring and do both!

It’s been a busy week as we get ready for a live demo at Homebrew Arcade.

We’re focusing on improving the visualization of combat. Now it’s easier to see where the shots are going with trails on the 3D map. We also implemented cover and missed shot handling so it’s been a productive week. Missed shots means if you miss the intended target, we still track the shot in case you hit something else interesting. 

This also reminds me to briefly mention how cover works. We showed in the video how when a target is obscured it is grayed out to make clear what’s in and out of your line of sight. If you fire, and hit the cover, it’s actually treated as armor. So you really want to hide behind something solid. A concrete wall is good cover, a cardboard box, not so much. 

Here is a short video clip showing “missed” shots being used to effect. I forgot to remove the debug logging where I was fixing some hit location code this week so ignore the text on the left.

Next week will be focused on AI.

Combat in Spy DNA

This is Jason with an update on the Spy DNA combat system.  We’ve been making some graphics and demo videos for our Kickstarter pitch, and I thought we should share some of them with you. In today’s post, I’ll start with our new gunsight, that we use for aiming, and then move on to the combat system.

In Spy DNA we have put a lot of thought into making combat feel as real as we can.  One of the areas that’s often disappointing in games is the critically important mechanism you use to attack the enemy. 

Most games assign a character a simple chance to hit, usually modified by range and cover.  We use a full 3D world instead.  To avoid the problems, such as identifying obstacles and cover, when aiming in the  top-down view, we open a gunsight view when you pick a target.  

Single shot firing sequence

The biggest thing we do differently is replace the “chance to hit” with “Circular error probable”.  The rings around your aim point represent the 50%, 90%, and 99% likelihood of your shot landing within those rings. 

You trade off time aiming for more accuracy in your shot.  You can adjust it to try to get the first shot off quickly, or take some time to make sure you hit. You can also adjust the number of shots, burst, or burst length.  

When using automatic fire, burst or full auto, recoil will reduce precision of later shots as recoil adds up.  For single shot and burst your character will re-aim so follow on shots meet the same accuracy requirement you set.

Our damage model is based on your weapon and where you hit the target, not on a random dice roll.  So you will be able to aim for weak spots in the armor or for vital parts of your target.  Cover is handled the same way.  The ability to move the target point around lets you aim for exposed parts of the target. 

Now let’s talk about how our combat system works. We call our system Concurrent Turn-Based.  I’ll explain what this means.  We differ from traditional turn-based games  in some important ways.  

Let me start with what we are trying to accomplish. 

1. The player should have time to think and take in the battlefield and environment.  
2. The moves available to the player should as close as possible mirror the options that  a real-life soldier would have.
3. The results of actions should be be realistic.

After a lot of experimenting we have settled on a system where the game focuses on a character when it is their turn to start their next action.  So while combat is ongoing, the game engine cycles though characters as their turns come up.  In this way it feels like a traditional turn-based game.  There is one very important difference.  While the game is progressing to the next player turn, every character and object in the game moves.  

This was not a decision we took lightly.  We made this decision to avoid the time quantization problem that traditional turn-based games have.  Think of the frustration where near the end of the player turn you move a character and trip one or more enemies.  Now your character (or whole party) just sits there helpless while the enemy takes a turn (or full round) worth of actions.  This is a side effect of games trying to map combat to a mechanism that doesn’t exist on a battlefield.  

In Spy DNA we are trying a more direct simulation of the world.  The character that makes contact would actually have the initiative.  The characters that spots them would make a reaction time roll (based on their attributes and combat experience) to see how fast they can react.  Also because other characters in the player party may be mid-action, such as movement, you could cancel those long actions and give them a new task.

I made a short video where a character ambushes two unaware NPCs.  The action commands I give the game are:

• Throw a grenade
• Draw my pistol
• Crouch
• Aim and shoot to finish off the second target

About three seconds of game time actually elapses in this demo.

Compared to Turn Based games, we have two major differences.  The first is turns in Spy DNA are not uniform in size.  Turns come up as the character completes their previous command.  This means that fast actions such as firing a single shot will result in that character's turn coming up again quickly.  Slow actions such as moving a long distance will mean many other characters are likely to take their turns before coming back to that character. 

The second major difference is the turns progress concurrently, i.e. all at the same time.  So if you give a move order to one character, and a quick attack order to another,  each time the second character attacks you will see the first make some progress on their move order.  In effect, you will see time progress forward for everyone until one of your characters completes all the commands in their queue.

Should a character spot something needing your attention while they completing an command, the game will stop and focus on the character.  This allows you to react to things that come up mid action such as an enemy coming around a corner.  

I hope this gives everyone a feel for the type of gameplay we are trying to deliver.  

Spy DNA gets a UI update

Jason from Shy Snake here with some screenshots from our UI update. We’ve been working through our UI, going screen by screen, to make it easier on the eye. While doing this, we have kept our focus on presenting information clearly to the player.

First up the attribute screen. This posed a bit of a challenge for us because we have a larger than normal list of attributes. We chose to have many attributes, to give each character a unique feel. For example rather than simply making a “strong” character you can be quick, powerful, or have great stamina.

Each primary attribute has three sub-attributes within it, to give the character extra detail. For the players that don’t want to see this level of information you can simply look at the major groups that give you an overview.

Next up the inventory screen. Here we went with a pretty standard list of icons with numbers to designate stacks of items. Selecting any item will fill the right side with a description of the item.

In this screenshot you can see some details on one of the games weapons. I’m taking this as an opportunity to show the attention to detail we put into the weapons in this game. While the values are not final you can get an idea of what we are building. 

At the bottom of the inventory screen you can see how we track encumbrance. In this case the character is lightly loaded so the effect is minimal. No significant effect on walking speed but a minor one to sprinting. We talked at length about the mechanics in an earlier post but here you can see it in action.

We've worked hard to make these changes, and the work isn't quite done yet. Tell us in the comments what you think of these updates!