NPC AI Survey, Optimization; & DLC 1 Level Deep-dive

NPC AI Survey, Optimization; & DLC 1 Level Deep-dive

Attention Officers,

Welcome to the 82nd installment of our Ready or Not Development Briefing, dated January 31, 2025.

Kicking off the first developer briefing of the new year, today we will introduce the new NPC AI sentiment survey, delve into a fresh approach to optimization, and explore in detail how we craft RoN levels, with illustrations from the Home Invasion DLC.

After returning from our winter hiatus, which concluded about a month ago, we are regrouping and intensifying our development efforts on Ready or Not. Currently, we are wrapping up a period of meticulous planning and prototyping.

The detailed explanation of our level design process in this briefing aims to shed light on just some of the work involved in each release of Ready or Not content. This aligns with the conclusion of our community mapping contest (please check our official Discord for additional information)!

On a related note, we are pleased to announce that some of the forthcoming levels, yet-to-be-announced, will be smaller in scale, which will help balance the grandiosity of recent massive levels (e.g., large levels like Leviathan, 3 Letter Triad).

Note: These development briefings are meant to inform you about portions of our ongoing support for Ready or Not. However, they do not cover everything we are working on at any given time. Please be aware that the content in our development briefings could still be undergoing changes and refinements.

NPC AI Sentiment Survey

As you may have already observed, we launched an NPC AI sentiment survey yesterday! We encourage you to complete it here:
https://forms.gle/DsVQT2x76qEpvWvj8

We are thrilled to have received an amazing number of responses already, and we would love to include yours—every piece of feedback is invaluable in helping us understand your experiences!

If you are uncertain about how to respond to a particular question, it likely is not mandatory, and you can bypass it! Should you inadvertently answer a required question without having an appropriate response, please utilize the written response sections to provide an explanation.

One of our enduring objectives for the game is the ongoing enhancement of SWAT AI, Suspect AI, and Civilian AI. This survey will aid us in collecting more precise data on the current state of AI following the release of Dark Waters, which incorporated bug fixes and improvements across multiple AI areas. We aim to analyze their impact and verify for any unintended issues.

A key focus of this survey is to dissect specific AI functionalities that influence the player’s perception of difficulty, as well as AI behaviors that may lead to either frustration or satisfaction. Although feedback from our social platforms and channels such as our official Discord is beneficial, this method allows us to pinpoint particular AI areas and consolidate that information in a singular location.

Optimization Endeavors

We are currently integrating the “Unreal Insights” system extensively into our ongoing performance optimization endeavors to ensure the game operates as seamlessly as possible. This system reveals performance metrics for specific game events and processes, enabling us to efficiently identify problem areas.

Already, we have witnessed significant early performance enhancements by incorporating this into our workflow, and as our technical artists and engineers increasingly utilize it, we anticipate even greater benefits.

[Image below: General example of the Unreal Insights task graph analysis view-port, from the UE5 Developer Page]

In-depth Exploration of Ready or Not Level Creation

Our dedicated level design team, in particular, Tisa for this briefing, documented their efforts in creating levels for Ready or Not and specifically for DLC 1.

This write-up illustrates our philosophies and processes for creating levels in Ready or Not, featuring new behind-the-scenes photos of the Narcos development and concluding with a comprehensive example from Lawmaker in DLC 1:

Level creation in Ready or Not follows a multi-stage process from idea conception to final polish.

[Image below: The levels highlighted (Narcos and Lawmaker) were inspired by feedback from our Home Invasion map poll]

Stage 1 (Planning)

The journey begins with production or deciding on which level to create. At VOID, every team member is encouraged to propose an idea they find compelling, always bearing in mind inspiration sourced from community feedback.

Everyone contributes their concepts in the form of a game design document (GDD), which is added to a catalog. From there, based on the DLC’s theme and its fun and innovative aspects, a decision is made on the level we will proceed to work on.

Our narrative designers and concept artists then start developing sketches for the level, and the lead level designer creates a design brief for each level designer to initiate the prototype pass.

[Image below: A sketch of the Narcos layout, highlighting doorways and other essential elements]

Stage 2 (Prototype)

The level design team embarks on the initial ‘prototype pass.’ Various iterations of the level are quickly crafted to investigate what aspects are engaging and how to approach developing the alpha version. Since Ready or Not is rooted in realism with intense themes, we strive to ground our levels in reality as well.

This involves extensive research, acquiring real-life references, visiting actual locations when possible, ensuring the architecture aligns with the designated level for this area, etc. It is not uncommon at this stage to abandon a level if it does not meet our standards. Prototype passes serve this particular purpose—like artists’ sketches, level designers create quick block-outs that can be discarded if they do not reach the quality we hope to achieve for our levels.

[Image below: Narcos layout based on research of similar neighborhood development design, resembling a GPS view]

Once individual level designers have crafted their visions, they are reviewed by our lead level designer. Together with the level designer responsible for that iteration, they determine which version to take forward to the Alpha pass.

Often, each prototype holds value, so the Alpha may integrate the best aspects from each prototype—a culmination of the finest elements.

Stage 3 (Alpha)

After selecting the prime prototype, we delve deeper into the block-outs, adding navigation, doors, AI, cover, and making the level playable. Once in robust shape, our combat designer steps in to fine-tune the AI for the desired combat dynamics. Afterward, our internal QA team begins thoroughly testing the level to address any significant issues at this phase.

[Image below: Blockout version of Narcos]

When we have a solid Alpha contender, each level designer writes a comprehensive level handover document, containing important information, screenshots, references, and the rationale for each part of the level. This documentation and the level itself are then handed over to the Environment artist assigned to that level. At this point, the remaining team members begin working on the level to implement all the features, such as the audio team recording necessary voice-overs, character art team creating the suspects and civilians, etc.

The level designer maintains close communication with all teams collaborating on the level to address any inquiries and ensure the design intent remains intact as it passes through other departments.

Stage 4 (Beta)

After completing the Environment pass and polishing the level, level designers return to finalize several aspects.

  • Collision pass – The level designer constructs blocking volumes to prevent players/AI from accessing non-playable areas, optimizing movement.
  • Visibility pass – The level designer adds visibility blockers to any geometric gaps to prevent AI from unfairly shooting players.
  • Navigation pass – At this juncture, level designers refine navigation, ensuring no navigation occurs in non-playable spaces. Our system relies heavily on navigation, encompassing spawns, cover, room clearance, etc., so this step is vital for optimization and avoiding AI bugs.
  • Door pass – The level designer verifies that all doors and their features function correctly for both the player and SWAT AI.
  • Cover pass – The level designer ensures all cover points are generated and correctly positioned for AI utilization.
  • Interactable pass – The level designer makes certain all custom logic continues to work and is accurately implemented.

[Image below: Blockout related to the Informant on Narcos]


[Image below: Red blockout triangles in Narcos illustrate sightlines from one doorway to another. The aim is for players to visualize each new doorway without excessive exposure to danger while navigating]

Stage 5 (Content Complete)

In this phase, the level designer collaborates with all other teams to ensure that all features for this level are successfully implemented and function as intended. The level undergoes rigorous testing by QA throughout this process to guarantee it is ready for public release.

DLC 1 Lawmaker Design Dev Commentary

Lawmaker Layout

With the Lawmaker level, we sought to craft a mansion featuring expansive rooms and broad corridors.
Most mansions boast unique layouts and peculiar corners absent in ordinary homes, adding a level of fun and distinction not yet seen in Ready or Not.

Despite the quirky angles and unique layouts found in mansions, they possess structural accuracy, often resulting in a coherent architectural plan when viewed from above.

The simplified plan for Lawmaker consisted of two rectangles intersecting at a 45-degree angle with a third rectangle. This design provided the elegant connecting corners and unique look we desired for the rooms and corridors, making the process of clearing rooms satisfying.

[Image below: Basic layout draft of Lawmaker]

To make the mansion easier to clear, we enlarged the rooms significantly compared to our usual RoN designs, taking full advantage of the increased space.

  • By expanding everything slightly, players and AI gain more maneuverability, reducing the likelihood of bumping into players or getting stuck in narrow passages.
  • Despite its massive appearance, the number of rooms in the mansion is consistent with our other maps, ensuring players are not overwhelmed with excessive door breaches, adhering to the Ready or Not formula.
  • The focus on larger rooms and hallways aligns with our goal for this level, as it offers realism and a challenge appreciated by our players.

[Image below: Lawmaker room segmentation]


Lawmaker Gameplay:
Before beginning creation, we contemplate the type of gameplay we want within these spaces. With Lawmaker, we envisioned a home invasion/hostage situation and wanted our AI systems to respond accordingly.

The intruders are a faction of eco-terrorists; not inherently formidable opponents. Except for two boss characters, the group lacks substantial weapon training and are somewhat astonished they managed to enter the house. Thus, behavior such as hesitancy, cowardice, and retreat is expected.

How do we convey this group as a threat and make them more challenging to apprehend?
Apart from an increase in hostage scenarios where caution is needed to avoid civilian casualties, we employed a zone system across multiple floors.

This means that suspects will engage players upon sighting them, and if given the chance, they will either attempt to flank the players or retreat to another zone, hoping to surprise the players again.
For Lawmaker, failure to apprehend them immediately would see them ascend a floor, leading to encounters with suspect groups on upper floors, echoing the next feature we wanted to integrate.

In reality, suspects often gather all civilians into a single room to hold them hostage.
We wanted to incorporate this scenario as well, making it critical for SWAT officers to be mindful of where they discharge their firearms in a room filled with both suspects and civilians.

With these considerations, we revisited the layout.
[Image below: Zone layout for Lawmaker]

We wanted the AI to exhibit flanking behavior, with the intention of most skirmishes occurring inside the house and minimal engagement in the exterior areas. Therefore, beside larger looping paths, we incorporated numerous mini loops within the house for the AI to exploit. This strategy offers substantial value to players who secure doors behind them or maintain watch over their backs.

As Lawmaker is a three-story mansion, we also considered flanking and AI zone retreating between multiple floors. Consequently, we ensured each wing and central section has a staircase connecting the floors.

These decisions enhance replayability, allowing players to select different paths during each playthrough based on the prevailing situation. To amplify replayability further, we provided players with three unique starting points for approaching the house, each presenting distinct challenges.

  • Main entrance offers the safest building access, placing players centrally, requiring choices on which side to clear first.
  • Side entrance grants access options through the center or rear of the west wing, though players are exposed to windows on the west wing during approach.
  • Back entrance provides the most entry possibilities, center, west, or east wing, but is highly exposed with risks from windows of all these sections.

Narrative \ Interactables

Following the storyline for Lawmaker, we wished to establish objectives and interactables to authenticate the narrative and foster player connection.

Once eco-terrorists invaded the mansion owned by Sven Anderson Lincoln, owner of the oil company they protested against, they severely vandalized the property, splattering oil and paint on priceless artworks, damaging sculptures, assaulting and capturing staff in the house, holding them hostage, etc. Sven’s wife and children retreated to a hidden panic room located on upper floors, while Sven was detained somewhere within the house.

Consequently, we sought to incorporate “Locate the family in the panic room” as one of our objectives. We developed logic for this event, making sure to address a few challenges:

  • The children are invulnerable during gameplay.
  • Multiple potential locations for panic rooms are set, varying with each playthrough.
  • We faced limitations in the available SWAT voicelines, considering this during development.
  • The active panic room needs to distinctly draw players’ attention upon discovery.

Initially, we placed the wife and children inside the panic room, with the expectation they remain until the SWAT team clears the entire property.
We then established a LIVE camera feed in the security room displaying other cameras around the property. Some were vandalized or destroyed, leaving the active panic room feed as the sole LIVE feed.

To ensure players received the correct information, we added an interactive button enabling players to switch all security monitors to the panic room feed and its location. Flashing red lights on active monitors were added to direct players’ attention to the security desk and interactive button, addressing the first challenge.

To address the second challenge, we placed two possible panic room locations randomly selected at match start. This is linked to our security room interaction, permitting players to identify the active panic room even before reaching it.

We overcame the third challenge by assigning dialogue to the mother upon player interaction with the panic room. Finally, red flashing lights were installed outside the panic room to signal its active status. They are positioned to be visible from hallways, preventing player frustration in locating them.

Another objective involved locating Sven, the mansion owner, within the house, adopting similar logic to our panic room system. By situating Sven in one of five different positions at map commencement, we ensure each play is unique due to his capacity for autonomous movement.

Conclusion

We highly encourage you to complete the NPC AI sentiment survey, as it greatly assists in shaping future enhancements to our AI systems! Lastly, a closing note on our level creation methodology:

In crafting each map, we strive to bring novel and intriguing elements into play—unique layouts, distinctive moods, innovative gameplay, formidable challenges, etc. Every level designer at VOID is genuinely passionate about this work, and we revel in witnessing how our players confront and navigate each challenge presented.

Despite comprehensive QA testing of our maps, we actively gather feedback from observing players’ interactions with each level. Should certain areas require modification, we are not hesitant to address these concerns in updates. Our ultimate objective is to craft a singular and enjoyable experience for our players.

This wraps up our 82nd development briefing—thank you for joining us. Stay tuned for more development insights in our next session!

Discover more about Ready or Not on Steam here.
Explore our other platforms: Discord, X, YouTube, Instagram, Facebook, Spotify.

Stack up and clear out.
VOID Interactive


Source