Skyscraper Elevator Optimization
← Project Index Skyscraper Elevator Optimization

Skyscraper Elevator Optimization

Analysis and optimization of high-rise elevator systems -; passenger routing and wait-time minimization -; framed as a SimTower-style vertical-city simulator where the dispatcher is the puzzle.

Archived Started: Summer 2025 Updated: Summer 2025

Overview

Anyone who's played SimTower remembers the moment the elevators stop being scenery and become the whole game -; red, furious little stick figures stacking up because your dispatch logic can't keep up. This project takes that moment seriously: a vertical-city simulator whose core challenge is high-rise elevator dispatch, where minimizing passenger wait time is the operations-research puzzle underneath the building game.

Background

It started as a SimTower/YootTower revisit -; the old Maxis "ant-farm" cross-section of a tower, where you place income-generating floors and then have to actually move people between them. The interesting part is that the elevator becomes a legible, real-time window onto a scheduling problem: passengers turn red when a car waits too long, so the quality of your dispatch algorithm is something you can literally see.

The intent was also pedagogical: an operations-research sandbox where the hidden dispatcher logic is exposed and editable -; queueing theory, scheduling, feedback loops, and a bit of economics, all falling out of one honest little system.

How It Works

The design centers on a deterministic, headless simulation engine (pure logic, no rendering) with a separate animated view layer -; Canvas/PixiJS, because hundreds of moving agents are too many for the DOM. The dispatch core is built on classic elevator scheduling heuristics:

  • SCAN / LOOK dispatch -; greedy, local-decision schedulers that sweep the shaft rather than try to see the future.
  • Collective-control zoning -; express cars dedicated to upper floors.
  • Stress as signal -; passenger frustration as a live readout of throughput vs. demand.

Tenant types, costs, adjacency synergies, and time-of-day rhythms live in data files, so the difficulty and the scenarios are tunable without touching the engine.

Current Status

Archived at the vision stage. The pillars, core loop, and progressive-disclosure unlock ladder are specified in design docs; the simulation engine itself wasn't built out before the project was shelved. It reads more as a worked-through design than a running game.

  • Design pillars and core loop locked in.
  • Architecture chosen: headless deterministic sim + Canvas/PixiJS view.
  • No implemented engine -; remains a spec.