(Small animated gif to showcase the general current state of the game)
Hi and welcome to the sixth development blog for Unleash!
This week has been focused on bug fixing and improving the stability of the game. The two main areas for bug fixing were networking, as we are in the process of updating our networking libraries, and the AI desynchronization issue that we have been aware of for a few weeks, but only now have had the time and opportunity to look into seriously. While bug fixing may not be particularly “sexy” to write or read about, it is of paramount importance for games so that the game runs with as few technical issues as possible. We are however also working on new graphical content, including a new fully upgraded Razorlegs model, new level design and leaderboards.
AI Desynchronization - No More!
Desynchronization issues in a deterministic game are generally horrible to debug and correct. There are so many things that can go wrong; the slightest floating point error or minor logical fallacy may result in completely desynchronized game states, which obviously breaks the gaming experience, when you and your mate both believe yourself to be the winner. We have known about a desynchronization issue with our AIs for a while, but we didn’t know why or where specifically the issue came from. Today we finally found the issue and luckily it was extremely simple to fix; it turned out to be a particular AI type (Mr. Think) accidentally using the local (human) player’s state, rather than its own. Since local players may have vastly different states (e.g. creature and structure state), this obviously lead to desynchronization. Now that the issue should be resolved, it is safe to play multiple humans vs bots again in Unleash!
Machine Learning AI
We have been working and investigating AIs based on machine learning for some weeks now. The prospects are phenomenal: Truly learning AIs that improve over time as more games are played, constantly learning from each other and real players how to play Unleash and win. However, the road ahead is still long before we see them actually playing any public games. The topic of machine learning AIs is huge and famous implementations include Google’s image recognition and autonomous vehicles, but we don’t know of any “real” (commercially viable) games that actually utilize machine learning, even though we think the prospects in this regard are absolutely fantastic. Hopefully, Unleash can start a new trend with totally awesome, always challenging and evolving AI in games.
We are working on leaderboards currently, using Steamworks functionality to achieve it. The idea is that each Tower Defense scenario will have its own leaderboard, and Line Tower Wars will of course also have one. Leaderboards will keep top 64 or possibly more globally, but we also want to show both your own rank and those immediately around your rank, in addition to your a leaderboard with just your Steam friends. Leaderboards are a part of an attempt at facilitating more competitive gameplay. While the implementation is fairly close to being ready, we don’t have any UI in place to show it yet. Thus, for now this is just a pre-teaser.
Our level design has been a bit of a pain point for us, for a while now. We feel like our levels are too simplistic and ‘bare’, rendering them a bit boring to look at. Therefore we have been discussing and prototyping solutions for level design. We have a few concerns in this regard: Levels should first and foremost be pleasing to look at, without taking away too much attention. The game is already quite colorful through our monsters, towers and visual effects. The level should not steal that attention away. Additionally, the level need to be low-budget on performance, so that they don’t cost too much precious processing power. Finally, it is our hope that we can release a ‘map editor’ eventually, enabling our players to design and create their own custom levels. These requirements obviously pose some interesting challenges that we are attempting to solve through prototyping.
(Screenshot of a work-in-progress for a possible new design of the lane separators)
New Razorlegs Tier 4 Model
We are working on a new version of the final evolution of the Razorlegs monster, which visually better reflects its function: Running directly towards the end of the lane at a relatively high speed. To that end, the idea is that it should have very long and spiky legs that can reach the ground, even as it moves above towers and walls. Additionally, we are experimenting with different ways for the animations to behave while the monster is above a structure.
(Current work-in-progress of new Razorlegs Tier 4 animated model)
This week saw a large amount of bug fixing and stability improvements, including a solution to the identified AI desynchronization. Additionally, we are working on new features such as leaderboards, smarter AIs and new levels, which may take several weeks before being ready for release.
We play Unleash every week, and you can join us for our play sessions. Next one is Tuesday the 7th of November at 17:00 / 5 PM (CET). Join us online for a multiplayer Unleash match!
If you want to see how we play, but cannot join the play session, you can watch gameplay videos without commentary through our Twitch profile. We live stream almost every session.
If you haven’t already, get your free Unleash beta Steam key through our website. The free open beta will run for a limited time period, so use the chance to try it out now for free with your friends!
And remember… Monsters never come alone.