Aerospace and Electronic Systems Magazine September 2016 - 41

OVERVIEW OF CONFLICT RESOLUTION STRATEGIES
The collision avoidance approaches can be briefly divided into six
categories: Rule-Based (RB), Force Field (FF), geometric, Numerical Optimization (NO), Markov Decision Process (MDP), and
Artificial Heuristic (AH), respectively.
The RB approach defines a conflict resolution as a pre-described maneuver for a specific type of observed potential conflict.
By prescribing an activation specified for each rule in the form of
position and orientation of airplanes, a set of rules for autonomous
separation are designed in [12]. Because the set of rules are fixed
for all airplanes in the shared airspace, it is not possible to integrate
further intentions into the conflict resolution process.
The FF method is formulated by assigning a potential function
to the free space, while the aircraft is recast as a particle reacting to
forces. The destination is modeled as an attractive force and the detected intruders in the airspace are regarded as repulsive forces instead. An analytically tractable potential field model of free space
is developed in [13], in which collision avoidance between the aircraft and threat can be guaranteed. FF provides a simple means to
generate a conflict resolution strategy. However, when intruders
are located close to each other, the existence of local minima can
trap the trajectory generation.
The geometric approaches use the geometry of relative kinematics between the UAV and the intruder to address path planning problems in the presence of the moving threat. By virtue of the current
position and velocity vectors, a geometric approach to the aircraft
conflict resolution is proposed in [14]. The three-dimensional airtraffic collision problem is formulated by a mixed-integer nonlinear

Figure 1.

The illustrative scheme of SAA functionality in a UAV.

SEPTEMBER 2016

programming problem in [15]. The optimal function for a UAV is
solved by minimizing total flight time while avoiding all possible
conflicts. The geometric approaches have been considered in many
cases to generate collision-free trajectories. However, all available
actions, including the change of heading, velocity, and altitude to
cover, lead to a very complex geometrical optimization problem.
The NO methods deploy a kinematic model of the aircraft together with a set of constraints and cost metrics. An optimal trajectory to resolve conflict is generated based on the lowest cost
and the most desired constraints. A randomized search method is
presented in [16] by applying a convex optimization-based cross
pattern to obtain a resolution maneuver with the minimum energy.
The conflict resolution in [17] is formulated as an optimal control
problem to minimize the flight time of airplanes flying in the horizontal plane with a constant speed. The limitation of NO methods
is that with increasing number of cost functions, the computation
burden is significantly increased.
Based on a stochastic model, MDP addresses the decision
problem by discretizing the system into a finite number of states
and actions [18]. By modeling the collision avoidance as a MDP,
the collision resolution logic is generated for the collision avoidance system by solving the MDP problem. To account for observation uncertainty, the Partially Observable Markov Decision
Process (POMDP) is introduced for modeling under uncertainty.
By constructing a MDP in [19], an optimal path is derived to successfully avoid multiple guns and reach the target. Reference [20]
develops a POMDP-based path planning for UAVs to track targets
while evading threats under wind disturbances. As an excellent approach of conflict resolution under uncertainty, POMDP guarantees UAV's safety in uncertain and dynamic environments. Nevertheless,
the disadvantage of POMDP is recognized as
the high computational complexity. The dimension of the POMDP-based solution grows
exponentially with the number of grids. As the
grid number is increasing, the convergence
rate slows down, while it becomes extremely
difficult to generate an optimal solution.
By mimicking the natural evolution process, AH approaches have been employed to
optimize a trajectory bypassing all intruders. Inspired by the social behaviors of bird
flocking and schooling of fishes, the Particle

IEEE A&E SYSTEMS MAGAZINE

Table of Contents for the Digital Edition of Aerospace and Electronic Systems Magazine September 2016