Fleet Rebalancing — Demand-Predictive Shuttle Repositioning

Fleet

Fleet size 8 shuttles

Rebalancing Strategy

Reposition interval 12 s

Demand intensity 1.0×

Live Metrics

Sim clock06:00

Avg wait time

— s

Coverage

—%

Idle / Busy

— / —

Reposition cost

—

Idle vehicle Repositioning (predictive) En route to pickup On trip with passenger

How it works

Each of the 36 grid zones has a time-varying demand rate built from two rush-hour Gaussian bumps plus a per-zone phase/amplitude offset (downtown zones peak at commute hours, residential zones peak in antiphase):

d_i(t) = base + A_i·exp(-((t-8)/w)^2) + A_i·exp(-((t-18)/w)^2)
P(new trip request in zone i during dt) = 1 - exp(-λ0 · k_demand · d_i(t) · dt)

Every idle or repositioning vehicle can be dispatched to a new pickup request; the nearest available vehicle is always chosen (minimises immediate travel distance, a standard greedy heuristic for online vehicle-request matching):

assign(request) = argmin_v∈idle∪repositioning ‖pos_v - zone_request‖

Reactive fleets only move once a request exists — idle vehicles just wait wherever their last trip ended. Predictive fleets additionally run a rebalancing pass every "reposition interval" seconds: each zone gets an uncovered-demand score

score_i = d_i(t) - k_cover · (idle vehicles already within radius R of zone i)
repositionTargets = zones sorted by score_i, descending

and idle vehicles are greedily driven toward the highest-scoring uncovered zones — pre-positioning supply where demand is predicted to peak next, at the cost of the driving distance shown in "Reposition cost". This mirrors the empty-vehicle repositioning problem operators of autonomous shuttle and micromobility fleets solve continuously: trade proactive driving cost against passenger wait time.

High demand zone Low demand zone

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.