Giannis P. Roussos

3D navigation and collision avoidance for a non-holonomic vehicle

This paper expands the methodology of Navigation Functions for the control of a spherical aircraft-like 3- dimensional nonholonomic vehicle. A Dipolar Navigation Function is used to generate a feasible, non-holonomic trajectory for the vehicle that leads from an arbitrary position to the target, in combination with a discontinuous feedback control law that steers the vehicle. The motion model used incorporates the nonholonomic constraints imposed on an aircraft, preventing any movement along the lateral or perpendicular axis, as well as preventing high yaw rotation rates. The control strategy provides guaranteed collision avoidance and convergence, and is supported by non-trivial simulation results.

Control of multiple non-holonomic air vehicles under wind uncertainty using Model Predictive Control and decentralized navigation functions

We present a novel control scheme for multiple non-holonomic vehicles under uncertainty, which can guarantee collision avoidance while complying with constraints imposed on the vehicles. Dipolar navigation functions are used for decentralized conflict-free control, while model predictive control is used in a centralized manner in order to ensure that the resulting trajectories remain feasible with respect to the constraints present and to optimize the performance objectives. The model used is chosen to resemble air traffic control problems, with some uncertainty introduced in the system. The efficiency of the control strategy is demonstrated by realistic simulations.

Decentralized and Prioritized Navigation and Collision Avoidance for Multiple Mobile Robots

We present an algorithm for the decentralised navigation of multiple mobile robots. Completely decentralised Navigation Functions are constructed, creating a potential field for each robot that gives rise to a feedback control law. The construction of the potential field incorporates limited sensing and explicit prioritisation in the form of priority classes. A non-circular sensing area creates asymmetrical sensing by reducing the influence of robots and obstacles behind each robot, introducing implicit priorities resembling “rules of the road”. Static and moving obstacles are also taken into account, as well as malfunctioning robots that are unable to maneuver. A decentralised feedback control law based on the gradient of the potential field ensures convergence and collision avoidance for all robots, while respecting a lower speed bound. Simulation results demonstrate the efficacy of the proposed algorithm.

Ground Assisted Conflict Resolution in Self-Separation Airspace

We propose a novel method for conflict resolution for aircraft flying in a self-separation airspace, with the possibility of some ground assistance. Our method is based on navigation functions, that can guarantee decentralized conflict avoidance for the short term horizon, supported by a model predictive controller on the ground, responsible for the optimization of the aircraft flight plans in the mid term horizon. As a consequence, our method combines the short term safety guarantees provided by navigation functions with the long term optimality and constraint satisfaction guarantees (in terms of airspeed, turning radius etc.) provided by model predictive controllers. The efficiency of the approach is demonstrated on simulations involving a number of aircraft converging in planar configurations.

Completely decentralised navigation of multiple unicycle agents with prioritisation and fault tolerance

We propose an algorithm for decentralised navigation of multiple independent agents, applicable to Robotics and Air Traffic Control (ATC). We present completely decentralised Navigation Functions that are used to build potential fields and consequently feedback control laws. Our approach employs local sensing, limited by a maximum sensing range and integrates priorities in the Navigation Function (NF) construction. Static and moving obstacles are taken into account, as well as agents that are unable to maneuver. A decentralised feedback control law is used, based on the gradient of the potential field, ensuring convergence and collision avoidance for all agents while respecting a lower velocity bound. An upper limit for the convergence time is given and simulation results are presented to demonstrate the efficacy of the proposed algorithm.

Towards constant velocity Navigation and collision avoidance for autonomous nonholonomic aircraft-like vehicles

This paper presents a methodology for the decentralised control of multiple 3-dimensional nonholonomic agents. The proposed control scheme is based on Navigation Functions and offers improvements compared to previous work of the authors in this field. The problem formulation is chosen to resemble Air-Traffic Management, where the safety guarantees provided by Navigation Functions based control strategies are very important. The linear velocity of each agent is maintained constant and equal to a desired value in most cases, while the azimuth and elevation control laws are engineered to reduce the required control effort. These qualitative improvements do not affect the collision avoidance characteristics of the control strategy. The performance and efficiency of our approach is supported by computer simulations presented in the paper.

Decentralised navigation and collision avoidance for aircraft in 3D space

We present an algorithm for the distributed navigation of nonholonomic aircraft-like agents in 3D space. The proposed control scheme offers improved applicability for aircraft navigation and compatibility with ATC practice wrt previous work. The algorithm maintains a desired horizontal velocity, while limiting the climb or descent angle within predefined bounds. Moreover, it is designed to favor straight and level flight, yielding more sensible manoeuvres that require reduced steering effort. Simulation results demonstrate the performance of the approach.

Decentralized Navigation and Conflict Avoidance for Aircraft in 3-D Space

We present an algorithm for the distributed navigation and conflict avoidance of nonholonomic aircraft-like agents in 3-D space. The proposed feedback control scheme offers improved applicability to aircraft navigation and compatibility with Air Traffic Management practice with respect to previous work. Our approach aims to maintain a desired horizontal velocity for each aircraft, while limiting the climb or descent angle within bounds according to aircraft performance characteristics. Moreover, the algorithm is designed to favor straight and level flight, resulting in more sensible manoeuvres that require reduced steering effort. The proposed control scheme is based on the Navigation Functions methodology and offers formally guaranteed conflict avoidance and convergence properties. The performance characteristics of our method are demonstrated through simulation results.

Completely decentralised Navigation Functions for agents with finite sensing regions with application in aircraft conflict resolution

We develop the framework of decentralised Navigation Functions for the case of multiple agents of arbitrary shapes and sensing areas around them moving in N-dimensional space. Unlike previous approaches utilising the Navigation Functions methodology, the construction here does not rely on diffeomorphisms, thus reducing the computational cost of the algorithm. The resulting potential field is absolutely locally computable and can be easily adapted to decentralised aircraft conflict resolution, as shown in the paper. Simulation results of multi-aircraft conflict resolution are included to support the efficacy of the complete control scheme.