Agentfly - Projects - Agent Technology Center

AGENTFLY is a software prototype of multi-agent technology deployment in aerial vehicles air traffic control supporting the free flight concept. All aerial assets in AGENTFLY are modeled as asset containers hosting multiple intelligent software agents. Each container is responsible for its own flight operation. The operation of each vehicle is specified by an unlimited number of time-specific, geographical waypoints. The operation is tentatively planned before take-off without consideration of possible collisions with other flying objects. During the flight performance, the software agents hosted by the asset containers detect possible collisions and engage in peer-to-peer negotiation aimed at sophisticated re-planning in order to avoid the collisions. The aim of this agent deployment is to demonstrate readiness of multi-agent technology for distributed, flexible, and collision-free coordination among heterogeneous, autonomous aerial assets (manned as well as unmanned) with a potential to (i) fly a higher number of aircrafts, (ii) decrease requirements for off-board control operators and (iii) allow a flexible combination of cooperative and non-cooperative collision avoidance.

AGENTFLY is build on top of the AGLOBE multi-agent platform. AGLOBE provides flexible middleware supporting seamless interaction among heterogeneous software, hardware and human actors. AGLOBE outperforms available multi-agent integration environments by its ability to model rich environments in which agents interact, by its support of full code migration and by its support for scalable experiments.

AGENTFLY provides:

distributed model of flight simulation and control,
time-constrained way-point flight planning algorithm avoiding specified no-flight zones,
flexible collision avoidance architecture, dynamic adjustment to changes in the flight environment
connectors to external data (Landsat images, airports monitors, no-flight zones, cities),
2D/3D visualization including a web-client access component, and
multiple operator – facilitating real-time control of selected assets

AGENTFLY provides four distinct collision avoidance (CA) algorithms linked by a flexible mechanism managing the autonomy of individual assets and selecting the best collision avoidance strategy in real time:

RULE-BASED CA ALGORITHM is a domain dependent algorithm based on the Visual Flight Rules defined by FAA2. Upon the collision threat detection, the collision type is determined on the basis of the angle between the direction vectors of the concerned aircrafts. Each collision type has a predefined fixed maneuver which is then applied in the replanning process. Visual flight rule-based changes to flight plans are done by both assets independently because the second asset detects the possible collision with the first asset from its point of view.
UTILITY-BASED CA ALGORITHM deploys multi-agent negotiation theories (namely Monotonic Concession Protocol with the Zeuthen Strategy) aimed at finding the optimal CA maneuver. The software agents on each asset generate a set of viable CA maneuvers and compute costs associated with each maneuver (based on e.g. the total length of the flight plan, time deviations for mission way-points, altitude changes, curvature, flight priority, fuel status, possible damage or type of load). The agents negotiate such a combination of maneuvers that minimizes their joint cost associated with avoiding the collision.
MULTI-PARTY CA ALGORITHM extends the above presented CA algorithm by allowing several assets to negotiate about collective CA avoidance maneuver. This algorithm minimizes the effects of CA maneuvers causing conflicts in future trajectories with other flying assets. While requiring more computational resources, this strategy has shown to provide more efficient free-flight collision free trajectories.
NON-COOPERATIVE CA ALGORITHM supports collision avoidance in the case when communication between aircrafts is not possible. Such a situation can arise e.g. when on-board communication devices are temporarily unavailable or when an asset avoids a hostile flying object. This algorithm is based on modeling/predicting the future airspace occupancy of the non-cooperative object and representing it in terms of dynamic no-flight zones. Based on this information, the algorithm performs continuous re-planning.