Unmanned Aerial Vehicles Evolutional Flight Route Planner Using the Potential Field Approach

Abstract

This study proposes a new flight route-planning technique for autonomous navigation of unmanned aerial vehicles (UAVs) based on the combination of evolutionary algorithms with virtual potential fields. By combining a radial force field with a swirling force field, three-dimensional virtual potential fields are constructed for repelling infeasible UAV flight routes from threatening zones. To ensure feasibility, major flight constraints are considered when searching for the optimal flight route. This study examines both single-and multiple-obstacle cases to determine the efficiency of the proposed flight route planner. The UAV navigation method uses an offline planner in known environments and an online planner for flight route replanning when popup threats emerge. Both planners were tested under various scenarios. The results show that the proposed planner can efficiently enable the safe navigation of UAVs. NomenclatureCroute k = objective term in a neural network system d k;i;j = distance of ith waypoint of kth flight route to jth obstacle froute k = constrained objective function in a neural network system F max = parameter used to modulate the potential field within R j and S j F Total k;i = resultant force influencing ith node of kth flight route F Multi k;i;j = virtual potential force F xy , F z = horizontal and virtual potential forces L max , L min = maximal and minimal lengths of route segment and amplitude of oscillation O xyz = virtual potential field O w i = swirling repulsive force o Multi j = coordinates of overall center mass Proute k = penalty term in a neural network system p n = end node; goal position p 0 = starting position p 1 = second node q f k;i = correction for ith node in kth chromosome according to magnitudes of virtual force R j = minimal radius measured from o j , which covers entire obstacle S j = sphere of influence (radial extent of force from o j ) s k;i = vector of ith segment in cylinder diameter of kth chromosome T 0 xyz = virtual force field of threat areas near dangerous zone v xy , v z = horizontal and vertical velocities x dir g , y dir g , z dir g = expected direction at goal position x dir s , y dir s , z dir s = direction of unmanned aerial vehicle at starting position max = maximal climbing angle of unmanned aerial vehicle max = maximal diving angles of unmanned aerial vehicle max = maximal turning angle of unmanned aerial vehicle x , y , z = direction of magnitude of mutation on corresponding axis

Real‐time route planning for low observable unmanned combat aerial vehicle

“…In (10), V and V are minimum and maximum flight speeds; in (11), 𝛼 is the angle of attack at which vehicle starts to stall; in (12), 𝜙 is the maximum bank angle; in (13), 𝜓 is the maximal change of yaw angle within a route planning cycle (such as t ′ + 1 → t ′ + 2) and it reflects the limit by vehicle's side overload. In (14), T(⋅) is the available maximal thrust associated with z and Mach number Ma (t), which is subject to (15); in (15), 𝜅 is adiabatic exponent, ℜ is real gas constant for air, Te (z) is the air temperature at altitude z. At altitude z, T(⋅) is represented as a quadratic function of Ma (t) in ( 16) 33 ; c 1 T (z), c 2 T (z) and c 3 T (z) are parameters with z. Initial/final condition for state/control variables: The components of state variable s (t) include x (t), y (t), V (t) and 𝜓(t); the components of control variable u (t) include the rest of variables except T t f .…”

“…Here to note that, this paper will not provide comparison analysis with the methods for UCAV's route planning in previous publications, since almost all these methods [8][9][10][11][12][13][14][15][16][17][18] did not consider (or over-simplified) the constraints of dynamics. If the route results do not fully respect the physical capability represented by the dynamics, they are in fact not flyable and thus have no meanings to make a direct comparison with the modeling integrated with the constraints of the dynamics.…”

“…The associated research efforts in previous publications focused on five principal approaches, which are A* method, 8,9 mixed integer and linear programming (MILP), 10,11 biology-based method, [12][13][14][15][16] indirect method [17][18][19] and direct method [20][21][22] :…”

“…The associated research efforts in previous publications focused on five principal approaches, which are A* method, 8,9 mixed integer and linear programming (MILP), 10,11 biology‐based method, 12–16 indirect method 17–19 and direct method 20–22 : the first class constructs digitized grid consisting of square cells to represent the region in which the route planning is performed. A “cost” value to every cell is prescribed to reveal the preference of UCAV traveling through the region, incurred by threat exposure or other factors.…”