Přemysl Volf

Agent-Based Cooperative Decentralized Airplane-Collision Avoidance

The efficiency of the current centralized air-traffic management is limited. A next-generation air transportation system should allow airplanes (manned and unmanned) to change their flight paths during the flight without approval from a centralized en route control. Such a scheme requires decentralized peer-to-peer conflict detection and collision-avoidance processes. In this paper, two cooperative (negotiation-based) conflict-resolution algorithms are presented: iterative peer-to-peer and multiparty algorithms. They are based on high-level flight-plan variations using evasion maneuvers. The algorithms work with a different level of coordination autonomy, respect realistic assumptions of imprecise flight execution (integrating required navigation performance), and work in real time, where the planning and plan-execution phases interleave. Both algorithms provide a resolution in a 4-D domain (3-D space and time). The proposed algorithms are evaluated experimentally, and their quality is studied in comparison with a state-of-the-art agent-based method-the satisficing game theory algorithm.

AGENTFLY: Towards Multi-Agent Technology in Free Flight Air Traffic Control

Ever rising deployment of Unmanned Aerial Assets (UAAs) in complex military and rescue operations require novel and innovative methods for intelligent planning and collision avoidance among a high number of heterogeneous, semi-trusted flying assets in well specified and constrained areas [1]. We have studied the free flight concept as an alternative to the classical, centralized traffic control. In free flight the unmanned aerial assets are provided with flight trajectory that has been elaborated without consideration of other flying objects that may occupy the same air space. The collision threads are detected by each of the aircraft individually and the collisions are avoided by an asset-to-asset negotiation. Multi-agent technology is very well suited as a technological platform for supporting the free-flight concept among the heterogeneous UAAs. In this chapter we present AGENTFLY, multi-agent system for free-flight simulation and flexible collision avoidance.

Distributed Platform for Large-Scale Agent-Based Simulations

We describe a distributed architecture for situated large-scale agent-based simulations with predominately local interactions. The approach, implemented in AglobeX Simulation platform, is based on a spatially partitioned simulated virtual environment and allocating a dedicated processing core to the environment simulation within each partition. In combination with dynamic load-balancing, such partitioning enables virtually unlimited scalability of the simulation platform. The approach has been used to extend the AgentFly air-traffic test-bed to support simulation of a complete civilian air-traffic touching National Air-Space of United States. Thorough evaluation of the system has been performed, confirming that it can scale up to a very high number of complex agents operating simultaneously (thousands of aircraft) and determining the impact of different configurations of the simulation architecture on its overall performance.

Surveillance of unmanned aerial vehicles using probability collectives

A rising deployment of unmanned aerial vehicles in complex environment operations requires advanced coordination and planning methods. We address the problem of multi-UAV-based area surveillance and collision avoidance. The surveillance problem contains non-linear components and non-linear constraints which makes the optimization problem a hard one. We propose discretization of the problem based on the definition of the points of interest and time steps to reduce its complexity. The objective function integrates both the area surveillance and collision avoidance sub-problems. The optimization task is solved using a probability collection solver that allows to distribute computation of the optimization. We have implemented the probability collective solver as a multi-agent simulation. The results show the approach can be used for this problem.

Multilevel approach to agent-based task allocation in transportation

We present a hybrid algorithm for distributed task allocation problem in a cooperative logistics domain. Our approach aims to achieve superior computational performance by combining the classic negotiation techniques and acquaintance models from agent technology field with methods from the operation research and AI planning. The algorithm is multi-stage and makes a clear separation between discreet planning that defines the tasks and allocation of resources to available tasks. Task allocation starts with centralized planning based on acquaintance model information that prepares a framework for efficient distributed negotiation. The subsequent distributed part of the task allocation process is parallel for all tasks and allows the agents to optimally allocate their resources to proposed tasks and to further optimize the allocation by negotiation with other agents. Parallel execution of the task allocation mechanism allows the algorithm to answer the planning request in predictable time, albeit at expense of possible non-optimality. In the experiments, we evaluate the relative importance of OR and negotiation parts of the task allocation process.

NAS-wide simulation of air traffic with ATC behavior model

The US National Airspace System (NAS) is incredibly complex, and consists of many specific functions. Given predicted increases in air traffic, enhancement of the current system and development of the NextGEN system are critical to maintain safe and efficient operation. Each feature of the system needs to be carefully designed, developed, tested and validated. Real-time human-in-the-loop (HITL) simulations represent one of the most powerful and realistic testing tools. HITL simulations can provide valuable feedback on how new features influence the behavior of human operators. The drawbacks of HITL simulations include limited flexibility and scalability, and high cost. Fast-time simulation is alternative option to HITL simulation, especially during research and initial development phases. Fast-time simulation provides reasonable detail of the simulation together with scalability and fast implementation of new tools. AgentFly system is presented as fast-time simulation suitable for NAS-wide simulation. It features cognitive behavioral model of Air Traffic Controller (ATC). AgentFly is used to perform several experiments with increasing air traffic to demonstrate suitability of fast-time simulation as a efficient tool for what-if analyses.