Show HN: Elevators

Elevator control turns out to be a surprisingly deep optimization problem, balancing wait times, total travel time, throughput, energy use, wear, and even equity between riders. Commenters react to an interactive explainer on classic algorithms like SCAN/LOOK and more advanced systems such as destination dispatch, noting that “smart” kiosks can underperform simple up/down buttons once real-world human behavior and the need for dynamic rebalancing are factored in. The thread branches into practical pain points (full cars stopping on every floor, bad parking strategies, rush-hour saturation), regulatory and safety constraints, and the long cultural tail of elevator sims and games like SimTower and Elevator Saga.

Algorithms & Optimization

  • Thread centers on classic algorithms (SCAN/elevator algorithm, LOOK, reservation-based schemes like RSR) and how they trade off average wait time, p90 wait, and throughput.
  • Some note that elevator scheduling resembles disk I/O scheduling and general “dial‑a‑ride” optimization problems.
  • Several commenters emphasize that throughput and capital cost usually dominate; minimizing cabs is often more important than shaving a few seconds off waits.
  • Others highlight that algorithms should consider non‑uniform traffic patterns (e.g., lobby↔floors dominates; near-floor trips may be rare due to stairs).

Destination Dispatch vs. Simple Buttons

  • Many report frustration with destination-dispatch kiosks: confusing UI, “locked‑in” car assignments, hard to reoptimize once riders are assigned.
  • Simulations described in the article suggest destination dispatch can underperform simple up/down buttons when real‑time rebalancing is disabled.
  • Some argue this is partly an artifact of the simulation; in real buildings with strong group patterns (office lunch groups, hotels, cruise ships), destination dispatch often feels faster and reduces in‑car stops.
  • There’s debate whether, in principle, “more information can’t hurt”; critics point out practical limits in communicating dynamic reassignments to humans.

Human Behavior & UX

  • A lot of perceived inefficiency stems from user behavior: pressing both up and down, jamming doors, not understanding destination dispatch, crowding into the wrong car.
  • People often optimize for “feeling of progress” (moving in the wrong direction vs. waiting still), not total time.
  • Psychological tricks (e.g., giving people something to do while waiting) matter as much as raw metrics.

Real-World Edge Cases

  • Conferences, furry conventions, and cruise ships produce extreme, highly directional loads that break usual assumptions; stories of 30+ minute waits are common.
  • Hotels and mixed‑use towers at peak times show failures when algorithms ignore load or allow full cars to keep stopping at every requested floor.

Hardware, Safety, and Constraints

  • Comments mention counterweights, regen braking, load‑sensing, and maintenance modes (e.g., exclusive move‑in service, fire service).
  • Building codes, regulatory regimes, and cost structures shape how many elevators get installed and which systems are viable.

Games, Education & Nostalgia

  • Many reference elevator‑scheduling games (Elevator Saga, SimTower, other sims) as fun ways to explore these algorithms.
  • Some recall using elevator scheduling as interview or classroom problems, highlighting its layered complexity.