CoFly 低空交通协同调度平台

From aircraft development to city-scale operations

An eVTOL is an electric aircraft capable of vertical take-off and landing.

UAM—Urban Air Mobility—uses eVTOL and related aircraft for commuting, connections, and logistics within cities and metropolitan regions.

BackgroundProject context

A single eVTOL developing into a city-scale UAM network — As more flight missions enter limited low-altitude airspace,UAM systems must coordinate flight status, weather changes, airspace rules, and multiple stakeholders.

Three urban air mobility challenges: dense airspace, changing environments, and multi-party coordination.

The operational challenge is shifting

The central question is moving from “how to fly” to “how to manage flight safely and efficiently under dynamic conditions.”

These changes show that low-altitude traffic requires a new approach to management.

Because “how to manage” is still too broad, we narrowed it through the regional context, coordinator tasks, and gaps in existing frameworks.

Problem Definition问题定义

The problem definition narrows through four layers: operational shift, regional context, coordinator tasks, and research gap. — How might AI help coordinators respond efficiently to weather-driven airspace changes while preserving their final control over complex, high-risk decisions?

From system functions to the coordinator’s judgement path

After defining the problem, we reviewed urban air mobility systems, task-processing models, and Human-AI coordination research to structure the coordinator’s cognitive path into four stages.

DevelopmentFramework

Framework mapping six urban air mobility services to a four-stage human information-processing model and a coordinator cognitive path.

Use automation for efficiency while preserving human judgement

The four-stage framework established the working sequence. We then translated it into three actionable design strategies.

Design StrategyDesign strategy

Dynamic geofencing

To address: Rapid weather changes make fixed geofences and blanket airspace closures too slow and imprecise for risk response.
Design approach: Update geofence status from weather risk and proactively identify affected airspace and flights.

Expected value

Finer airspace control
Earlier risk identification

Cognitive-stage interaction

To address: Function-led interfaces scatter information and force coordinators to integrate it manually across modules.
Design approach: Organize information and actions around see, understand, decide, and track to keep judgement continuous.

Expected value

Shorter decision path
Less cognitive switching

Human-AI roles in exception flows

To address: Alert-only systems leave complex conflict resolution entirely to human operators.
Design approach: AI generates options, the coordinator confirms or adjusts them, and the system executes while reporting outcomes.

Expected value

Faster response
Final authority stays human

AI does not replace the coordinator. It handles frequent information processing and option generation so people can focus on complex judgement.

Clarifying core Human-AI interactions in the dispatch workflow

After defining the strategies, we assigned Human-AI responsibilities in the dispatch workflow: the system identifies the conflict type, then either handles it automatically or requests human intervention.

At the system level, the loop has three parts: data input, system processing, and execution with feedback.

Core interactionHuman-AI Workflow

The system loop connects data input, system processing, and execution with feedback.

CoFly Human-AI dispatch platform

CoFly Human-AI workflow for routine and exceptional conflicts across four cognitive stages. — Automation changes with risk and uncertainty:the more standard the scenario, the more AI handles; the higher the risk, the deeper the coordinator intervenes.

Let the interface follow the coordinator’s judgement sequence

We mapped the path from seeing risk and understanding conflict to decision-making and execution tracking into the interface layout.

Interface Layout页面布局

CoFly low-fidelity dispatch interface with an AI collaboration zone, central real-time airspace map, human decision zone, and execution timeline.

Layout iteration

In the first low-fidelity layout, flight and weather information sat below the map. Coordinators had to switch their gaze and combine information manually, weakening the map as the primary situational-awareness view.

We therefore reorganized the information hierarchy: