Decentralised navigation and collision avoidance for aircraft in 3D space

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

“…In fact, any controller that can maintain a decreasing rate for each potential Φ i , i.e.Φ i < 0 can be employed in combination with the potential field presented previously to stabilise the agents to their targets, while avoiding collisions. Such a control law has been presented in [8] for unicycle aircraft-like agents in 3D space.…”

“…We propose an absolutely locally computable potential field that takes into account multiple agents according to their priorities, as well as static and moving obstacles. This potential field used in a control scheme such as the one presented in [8] can offer decentralised, non-cooperative navigation for multiple agents. In fact, any controller that ensures a decreasing rate for the potential's value over time is applicable.…”

“…Deriving a controller for planar unicycles using the same principles can be achieved by neglecting the vertical velocity in [8]. The resulting control scheme employs the projection of the gradient ∇ i Φ i = Φ ix Φ iy ⊤ on the agent's i longitudinal (heading) direction:…”

“…In this paper we further develop limited sensing in the NF methodology, which combined with a feedback control law designed on the principles of [8] yields a completely decentralised solution for multi-agent navigation and collision avoidance in a workspace with obstacles. Our approach requires no a-priori computation or knowledge and does not rely at all on centralised controllers.…”

“…The main advantages of this class of methods are the formal performance guarantees they can provide, computational efficiency and their real-time feedback nature, that can compensate for measuring and modeling errors. Navigation Functions have been so far applied of multi-agent problems ranging from robotic navigation [7] to ATC applications [8].…”