November 22, 2022


Illustration of the three competing concepts developed in Metropolis 2. From left to right: centralized, hybrid, distributed

Centralized deconfliction is done before the take-off. The UAV operations are (pre-)planned in the way such that, as long as drones follow their , at any moment of time the distance between any pair of drones will be greater than the minimum allowed, meeting robust separation criteria and eliminating the need to invoke UAVs onboard equipment.  

As input to the deconfliction algorithm we have the operations demand (origin–destination pairs and intended departure time) for some time horizon T. The output are drones’ flight plans. A flight plan is a plan for the 4-dimensional trajectory (4DT) of the drone, i.e., a specification of its xyz coordinates for each time t.  

The main features of the centralized concept are that we have full knowledge about all the demand and that moreover, the knowledge is collected in a single place, allowing one to compute the flight plans together. The main feature of the output is that the flight plans are conflict-free.

In our implementation of centralized deconfliction, a flight plan is a polygonal path in a fixed layer (no layer changes on route). It is assumed that the vehicle always fly at their nominal (maximum operating speed) speed unless it is restricted by the geovectoring or turning speed restriction (ground speed is further affected by the wind Thus, the 4D trajectory for the drone is defined as soon as the time when the drone starts following the flightpath is specified. VTOL segments are added at the endpoints of the flight plan at the origin and the destination.

Airspace structure

  • Drone flies on predefined cruising layers => facilitates separation
  • Cruising layers separated by 30ft => 16 layers
  • Climb/Descent only at vertiport nodes => single cruising layer per drone
  • City is divided in two zones: open and constrained
  • Constrained airspace with street network
    • modelled as a graph
    • bi-directional streets
  • Open airspace is discretized 
    • using hex-grids
    • modelled as a graph 

The Hybrid concept consists of three main separation management components:

  1. Airspace structure
  2. Centralized pre-flight strategic separation management
  3. Decentralized in-flight tactical separation management

The Hybrid concept assumes that preventing conflicts is better than resolving conflicts. Therefore, the first two components, namely the airspace structure and the centralized pre-flight strategic separation management components aim to prevent as many conflicts as possible by ensuring that drone flight plans are largely conflict-free. But because of uncertainties such as wind, conflicts may still occur during flight execution. The decentralized in-flight tactical separation management component addresses such conflicts.

The following figure describes the relationship between the main components of the Hybrid concept. Here it should be noted that the strategic and tactical algorithms resolve conflicts along different dimensions: the strategic algorithm primarily uses horizontal resolutions while the tactical algorithm primarily uses vertical resolutions. In this way, the tactical algorithms aim to not disturb the decisions made by the strategic algorithm. Both algorithms take into account the constraints posed by the airspace structure when planning routes and computing resolutions.

Airspace Structure

  • Cruising layer: these are used by the strategic separation management algorithm to plan the route of a drone from origin to destination.
  • Resolution layers: these are (mainly) used for tactical deconfliction purposes.
  • Climbing/descending only at nodes
  • Speed limits in the cruising layers to reduce relative velocities
  • No speed limits in the resolution layers => hovering possible
  • Separate layer just for emergency drones => reduce interactions with ‘other’ traffic

The decentralised concept seeks to allow individual agents as much freedom as possible in order to accommodate the massive numbers of predicted urban air traffic. The consequence is that the separation responsibility lies with the individual agent. The decentralised concept will, however, also have a central authority to provide information (such as traffic) and flow constraints (implemented using geovector limits) for airspace areas. This central authority can be seen as a combination of centrally imposed dynamic road traffic speed limits and traffic information provided by Google Maps: it tries to prevent further congestion by imposing geovector limits in dense areas, and also provides traffic information to other drones for path finding purposes. The different components considered within the decentralised concept are presented in the following figure.

Airspace Structure

  • Streets are uni-directional
  • Street graph GA optimised for:
    • Avoiding same-level converging streets
    • Maximising optimal-distance connectivity
  • Layers of 30 feet height
  • Each layer corresponds to a range of heading values 

The concepts in practice

Decentralised concept

Hybrid concept

Centralised concept

Scroll Up