#Screeps ai code

#Screeps ai code

It had some leftover walls blocking some of our planned room sites, so I added some code to destroy or dismantle structures that are impeding planned structures. First, there was only one room available that met our criteria, at the very edge of a LawyerвЂ™s range. When I pushed the code up to MMO, I ran into a couple snags. I rearchitected a few things to make development easier, but there were only a couple multi-room bugs. Tests in my private server went surprisingly well. So, Build requests can now harvest energy from Sources as well as withdrawing from a logistics store. This requires resources from the parent office, but we donвЂ™t want to disrupt the work weвЂ™ve put into stabilizing the energy pipeline. Slightly more complex is the initial bootstrapping required to build the first spawn. Thanks to the Behavior abstractions we previously implemented, this was simple to add. Once reserved, a new Office will be instantiated. The office will spawn a Lawyer, dispatch it to the target controller, and attempt to reserve it. WeвЂ™ll create a new Acquire request that will be handled by an officeвЂ™s LegalManager. If weвЂ™ve scouted the room and determined a viable room plan (which includes things like checking for two sources, swampiness, etc.) then itвЂ™s a candidate for expansion. ACQUIRE is what weвЂ™re interested in here. If ( ! controller ) Д®XPLOIT will be reserved for future remote mining efforts. WeвЂ™ll start with a simple formula to set our intention for each territory. Our WorldState models are caching some of the relevant details even when we donвЂ™t have vision, so this gets us all the raw information we need. Territory IntelligenceД¬urrently we are sending an Intern out to scout nearby territories. Once the spawn is up and running, the Office can become self-sustaining. WeвЂ™ll need some way to request these minions from another Office. WeвЂ™ll send a Lawyer to claim the controller.ДЇinally, the Office will need help to get started - at least an Engineer to build the Spawn and a Paralegal to sustain the controller. Second, we need to initialize the new Office. If weвЂ™re doing remote mining eventually, weвЂ™ll want to plan ahead for that and make sure there are good remote rooms nearby. We want to colonize two-source rooms with few swamps and where we can fit a complete room plan. LetвЂ™s break this into a few steps.ДЇirst, we need to evaluate potential sites. It is time to expand! We are at GCе…ѓ on the MMO server and our private server test scenarios are easily reaching GCL2. Hauling energy from sources to a central distribution point is more efficient than hauling from sources to fulfill requests directly.You should be able to hit RCL 3 + tower before your initial safemode runs out.Have some kind of room planning, even if itвЂ™s just a basic stamp.Try to predict rather than react, especially when deciding minions to spawn.Creeps with specific roles are more efficient than general-purpose creeps, even at RCL1.Specific to building a reliable energy pipeline in Screeps: I defaulted to classes for my Managers and Analysts, in order to implement lifecycle functions, but let them grow out of control, resulting in a lot of unnecessary boilerplate elsewhere in my code. For one example, check out вЂњArtificial Intelligence for GamesвЂќ. They exist to make your development easier. If you have to fix an issue, you can do it in one place instead of ten (more often than not). Break out common snippets (like telling a creep to get energy) into clearly-named functions. Whether you log to the console, write to Grafana, or display dashboards with screeps-viz, insight into what your code is doing is invaluable for debugging and prioritizing improvements. But so far, IвЂ™ve collected some important lessons (and mistakes). And we still have a long way yet to go - we havenвЂ™t even touched terminals, labs, power, or advanced combat, among other things. Nineteen posts in! This has gone a lot longer than I expected when I started almost a year ago. HereвЂ™s the GitHub repo if youвЂ™d like to follow along. The article below describes the AI in its current state: IвЂ™m still expanding and refining my codebase.