betaFeedback

WorksComputer Science

Mass Scaling Laws for Energy-Optimal Horizontal Multirotor Flight

Abstract

Multicopters are a type of aerial robot seeing extensive real-world applications but with range and endurance as critical bottlenecks. Optimization of its energy efficiency through energy-aware flight planning (e.g. cruise speed-scheduling) has hence been an important topic. The mass of the multicopter is an important factor to consider, as it has a significant impact on efficiency, and many common applications are subject to frequent mass variation due to payload change. This work derives closed-form scaling relationships between the energy-optimal steady horizontal flight behavior of a multicopter and its vehicle mass, and validates them experimentally on two multirotor platforms. Specifically, for mass-only variation, i.e., when payload changes the vehicle mass while the airframe geometry, aerodynamic characteristics, and rotor configuration remain unchanged, we show through a combination of numerical simulation, theoretical derivation, and experimental validation that (1) the energy-optimal velocity is proportional to the square root of vehicle mass; (2) the energy-optimal pitch angle is constant with respect to mass; and (3) the optimal energy per meter is proportional to mass. The analysis shows that these three relationships arise from a common mass-similarity property of the steady-state flight and rotor-aerodynamic equations. Furthermore, the obtained scaling relationships enable simple yet effective cruise-speed planning to optimize energy efficiency and extend the range and endurance of the multicopter.

Introduction

MULTICOPTERS have been emerging as a versatile aerial robot with wide applications, ranging from aerial photography and inspections, to intelligence, surveillance, and reconnaissance (ISR) missions, to package delivery among many others. However, their widespread adoption faces a major challenge in range and endurance, as their flight time, typically around 30 minutes, poses significant constraints on their capability and performance [1].

Prediction and optimization of multicopter energy consumption and performance have been actively studied in aerial robotics. Extensive work has focused on modeling energy dynamics and estimating flight range and endurance [1, 2, 3, 4, 5, 6, 7]. While many early models focus on hovering or quasi-hovering conditions, recent studies consider the coupled rotor aerodynamics, propulsion, and rigid-body dynamics during forward flight [1], [8]. These aerodynamic effects make energy consumption strongly dependent on horizontal cruising velocity, with competing effects such as rotor inflow and body drag giving rise to an energy-optimal velocity. In [9], experimental measurements are performed to quantify the difference in energy consumption caused by horizontal velocity, while several model-based studies, including [1], [8], [10] among others, investigate the relationship using simulation. These works reveal the trade-off of various factors affecting the energy consumption rate, such as inflow aerodynamic effects and body drags, and the existence of an optimal velocity that maximizes the efficiency. Therefore, cruise-speed selection has become an important topic in energy-optimal flight planning and control [1], [9], [11, 12, 13].

One remaining topic of significant theoretical and practical interest is how energy efficiency and energy-optimal flight behavior change with vehicle mass. In many multicopter applications, e.g. drone delivery and air mobility, vehicle mass is subject to frequent and substantial variations, making its impact on energy efficiency important. Several existing works [10], [14, 15, 16] have investigated how the energy consumption versus velocity curve changes with mass through simulations and experiments. Beyond the expected increase in energy consumption, these studies show that the energy-optimal velocity also increases with mass when other vehicle properties, including aerodynamics and geometry, remain

(Corresponding author: Xinfan Lin). unchanged. The form of this dependence, however, has not been established. In [14], it appears as a qualitative trend in a numerical endurance study as the payload-to-vehicle weight ratio is varied, with no explicit relationship stated. In [10], the optimal speed is obtained numerically from an energy model calibrated to a reference vehicle, in which vehicle weight enters only the induced-power term while rotor speed, and hence blade profile power, is held fixed. Neither work identifies the invariance of the energy-optimal pitch angle, from which the velocity scaling follows, or addresses the scaling of the minimum energy per meter. Existing analytical models have also characterized the relationship among energy consumption, velocity, and vehicle mass. A commonly used metric is Energy Per Meter (EPM), i.e., the energy consumed per unit distance. For example, [17] extends a hovering-based energy model to horizontal flight to study the metric, while [18] derives velocity-dependent power consumption for forward flight and obtains energy per distance by dividing power by velocity. However, the former neglects forward-flight aerodynamic effects, while the latter simplifies them through a prescribed lift-to-drag relationship. In summary, the quantitative relationships between vehicle mass and energy-optimal flight conditions remain to be established.

The contribution of this work is an in-depth theoretical, numerical, and experimental study of the relationships among energy efficiency, cruise velocity, and vehicle mass, as summarized in Figure 1. Building on established multicopter energy dynamics [1, 8, 19], we combine a propeller model based on blade-element/momentum theory (BEMT) with steady-state force balance from the rigid-body dynamics to derive explicit mass-scaling relationships for the energy-optimal cruise condition. Specifically, for mass-only variation of a fixed vehicle configuration, i.e., when payload changes the vehicle mass while the airframe geometry, aerodynamic characteristics, and rotor configuration remain unchanged, we show that (1) the energy-optimal velocity is proportional to the square root of the mass; (2) the energy-optimal pitch angle remains unchanged with mass; and (3) the minimum energy consumption per unit distance, or EPM, is proportional to the vehicle mass. These relationships are derived theoretically and validated experimentally on two multirotor platforms under payload-induced mass variations.

Energy-optimal cruise behavior of a multicopter under payload-induced mass variation

Figure 1. Energy-optimal cruise behavior of a multicopter under payload-induced mass variation. (a) Experimental octocopter platform used for outdoor flight tests, shown on the ground (left) and during steady horizontal flight (right). (b) Theoretical energy consumption per unit distance normalized by mass (EPM/mm) versus forward velocity for two vehicle mass configurations, demonstrating collapse of the mass-normalized energy efficiency at the energy-optimal operating point (marked). (c) Scaling laws governing energy-optimal steady horizontal flight under mass variation for a multicopter: the optimal pitch angle is invariant to mass (Θ∗∝const.\Theta^* \propto\mathrm{const.}), the optimal cruise velocity scales with the square root of mass (Vx∗∝mV_x^* \propto\sqrt{m}), and the minimum energy consumption per distance scales linearly with mass (EPM∗∝m\mathrm{EPM}^* \propto m).

Multicopter Energy Dynamics Modeling

The theoretical and simulation studies in this work are based on a physics-based multicopter model that captures key coupled dynamics governing energy consumption, with a particular focus on rotor aerodynamics and rigid-body dynamics. Strictly speaking, the analysis targets the mechanical and aerodynamic efficiency, which constitute the dominant contributors to energy consumption and are essential for characterizing energy-optimal cruise behavior. Other system components, including the motor, battery, and electronic speed controller (ESC), are not considered explicitly in the closed-form derivations, as their efficiencies are relatively constant over the operating regime of interest and have a comparatively smaller impact on the mass-scaling behavior. The influence of these components is discussed separately in Sec. VI-B.

Rigid Body Kinetics

The motion of the multicopter is modeled using the rigid body dynamics according to the free body diagram in Figure 2. We consider 2D dynamics here, which are sufficient for forward flight,

Multicopter Free Body Diagram

Figure 2. Multicopter Free Body Diagram

X¨=∑j=1NpTjsin⁡(Θ)m−CBDVx2m\ddot{X} = \sum_{j=1}^{N_p} \frac{T_j \sin(\Theta)}{m} - \frac{C_{BD} V_x^2}{m}
Z¨=∑j=1NpTjcos⁡(Θ)m−g\ddot{Z} = \sum_{j=1}^{N_p} \frac{T_j \cos(\Theta)}{m} - g
Θ¨=τΘJΘ=∑j=1NpLΘ,jTjJΘ(1)\ddot{\Theta} = \frac{\tau_{\Theta}}{J_{\Theta}} = \sum_{j=1}^{N_p} \frac{L_{\Theta,j}T_j}{J_{\Theta}} \tag*{(1)}

where XX and ZZ are the horizontal and vertical positions, VxV_x and VzV_z are the horizontal and vertical velocities of the vehicle (center of mass) in the global frame, and Θ\Theta is the pitch angle. In addition, ∑Tj\sum T_j is the sum of thrusts of all rotors computed by the propeller model (with each rotor indexed by subscript jj and total number of rotors NpN_p), CBDVx2C_{BD}V_x^2 is the body drag force with CBDC_{BD} being the body drag coefficient, τΘ\tau_{\Theta} is the total pitch-axis torque on the vehicle generated by the thrusts, JΘJ_{\Theta} is the moment of inertia about the pitch axis, and LΘ,jL_{\Theta,j} is the arm length of each rotor thrust to the pitch axis.

Noted that we model the horizontal aerodynamic resistance as a quadratic body-drag term, and omit the linear drag component associated with rotor hub and in-plane forces, which are known from blade-element theory to grow roughly in proportion to the forward velocity [20]. For the flight regime of interest, i.e., typical mid- to large-scale payload-carrying multirotors operating near energy-optimal velocity, the linear drag is comparatively small. Aerodynamically, these vehicles operate at low advance ratio because the rotor-tip speed is much higher than the forward velocity, and hence the rotor-generated in-plane forces scale approximately linearly with velocity. In contrast, the airframe and payload behave as bluff bodies whose drag grows with the square of velocity [21]. A drag law that increases faster than linearly is needed for a pronounced non-monotonic EPM-velocity curve typically observed in experiments. We have performed experimental studies to validate such treatment. Specifically, we compute the rotor thrust and linear drag for steady forward-flight conditions using a blade-element/momentum model calibrated on thrust-stand data reported in literature [8]. Then, using real-world flight data, we determine the total drag directly from the measured velocity and pitch angle through steady force balance. The body drag term, which is the residual between the total and linear drag, grows approximately with the square of velocity in the neighborhood of the energy optima, and clearly dominates over the linear drag contribution in that regime. These physical considerations and experiment results together justify representing the dominant horizontal resistance near the optimal velocity as a quadratic drag term.

Finally, the velocities of each rotor (center) in the vehicle frame can be calculated as

vx,j=Vxcos⁡(Θ)+Vzsin⁡(Θ)vz,j=−Vxsin⁡(Θ)+Vzcos⁡(Θ)+Θ˙xj(2)\begin{aligned} v_{x,j} &= V_x\cos(\Theta) + V_z\sin(\Theta)\\ v_{z,j} &= -V_x\sin(\Theta) + V_z\cos(\Theta) + \dot{\Theta}x_j \tag*{(2)} \end{aligned}

where xjx_j is the x-axis position of the propeller relative to the vehicle center of mass.

Propeller Aerodynamics

Based on the blade element and momentum theory, the propeller aerodynamic model computes the thrust TrT_r and torque QrQ_r generated by a propeller (the subscript jj is omitted in this subsection as all propellers are the same during horizontal flight). This is achieved by integrating dTdT and dQdQ of propeller blade elements, of which the cross section is shown in Figure 3a, along the blade length coordinate rr and averaged over one cycle of the blade rotation angle ψ\psi,

Geometries of a Blade Element Cross-section and Air Flow Momentum of Propeller Disk

Figure 3. (a) Geometries of a Blade Element Cross-section and (b) Air Flow Momentum of Propeller Disk

Tr=∫02π∫R0RdT2π dψdr,dT=Nbρupl2ca(θ−uprupl)2Qr=∫02π∫R0RdQ2π dψdr,dQ=Nbρrupl2c(ϕa(θ−uprupl)+cd)2(3)\begin{aligned} T_r &= \int_{0}^{2\pi}\int_{R_0}^{R}\frac{dT}{2\pi}\,d\psi dr,\qquad dT = \frac{N_b\rho u_{\mathrm{pl}}^2c a\left(\theta-\frac{u_{\mathrm{pr}}}{u_{\mathrm{pl}}}\right)}{2}\\ Q_r &= \int_{0}^{2\pi}\int_{R_0}^{R}\frac{dQ}{2\pi}\,d\psi dr,\qquad dQ = \frac{N_b\rho r u_{\mathrm{pl}}^2c\left(\phi a\left(\theta-\frac{u_{\mathrm{pr}}}{u_{\mathrm{pl}}}\right)+c_d\right)}{2} \tag*{(3)} \end{aligned}

In the equation, ρ\rho is the air density, NbN_b is the number of blades per propeller, cdc_d is the blade drag coefficient, aa is the blade lift coefficient factor, cc is the blade chord length, θ\theta is the twist angle of the blade, and RR and R0R_0 are the total and base length of the blade. Full details of the derivation with underlying assumptions are available in literature [1, 8]. The main idea of Eqn. (3) is to determine the forces based on the dynamic properties of the air inflow to the blade, including the planar inflow velocity along the blade uplu_{\mathrm{pl}}, perpendicular inflow velocity upru_{\mathrm{pr}} vertical to the blade, and inflow angle ϕ\phi, which are shown in Figure 3a. These variables can be computed based on the various components of the relative velocity of the air to the blade, including the rotation and downwash of the propeller and motion of the whole multicopter, as

upl(r,ψ)=ωr+vxsin⁡(ψ),upr(r)=vi−vzu_{\mathrm{pl}}(r,\psi)=\omega r+v_x\sin(\psi),\qquad u_{\mathrm{pr}}(r)=v_i-v_z
ϕ(r,ψ)=tan⁡−1(upr/upl).(4)\phi(r,\psi)=\tan^{-1}(u_{\mathrm{pr}}/u_{\mathrm{pl}).} \tag*{(4)}

Since the vehicle-frame zz axis points upward, so vi−vz=vi+Vxsin⁡Θv_i-v_z=v_i+V_x\sin\Theta is the total inflow through the rotor disc: the freestream and induced components pass through the disc in the same direction and therefore add. ω\omega is the angular velocity of the propeller rotation. The downwash effect is captured by the induced air velocity viv_i of the rotor disk, which is the downward velocity of the air imparted by the propeller. This velocity needs to be determined by applying the momentum theory to the propeller air stream. According to Figure 3b, air enters the propeller airflow at location 1 with a relative velocity vv equal to the total UAV airspeed,

v2=vx2+vz2.(5)v^2=v_x^2+v_z^2. \tag*{(5)}

The airflow is accelerated perpendicularly by the propeller disk plane, resulting in an induced velocity viv_i at the disk (location 2) and vov_o at the stream outlet (location 3), both assumed to be uniform across relevant cross-sections. By applying the principles of momentum and kinetic energy, vov_o is determined to be twice of viv_i. Subsequently, the relationship between the thrust TT and the induced velocity viv_i can be established through the momentum principle [22],

Tr=m˙airvo=2m˙airvi(6)T_r=\dot{m}_{air}v_o=2\dot{m}_{air}v_i \tag*{(6)}

where m˙air\dot{m}_{air} is the air mass flow rate at the propeller,

m˙air=ρπR2vx2+(vi−vz)2.(7)\dot{m}_{air}=\rho\pi R^2\sqrt{v_x^2+(v_i-v_z)^2}. \tag*{(7)}

Furthermore, by substituting vxv_x and vzv_z from the rigid body dynamics relationship in Eqn. (2) and applying the steady-state horizontal flight conditions with Vz=Θ˙=0V_z=\dot{\Theta}=0, a quartic equation is obtained for solving viv_i

vi4+2Vxsin⁡Θvi3+Vx2vi2=(Tr2ρπR2)2,(8)v_i^4+2V_x\sin\Theta v_i^3+V_x^2v_i^2=\left(\frac{T_r}{2\rho\pi R^2}\right)^2, \tag*{(8)}

where TrT_r is the thrust of a single rotor, so that the total thrust is T=NpTrT=N_pT_r. Since all rotors are identical and equally loaded in steady horizontal flight, TrT_r is the same for every rotor.

Energy and Power Consumption

Power consumption of the multicopter is computed as the sum of the mechanical power of each propeller

P=∑jNpQr,jωj.(9)P=\sum_{j}^{N_p}Q_{r,j}\omega_j. \tag*{(9)}

which is integrated over time to obtain energy consumption. Since all propellers operate identically in steady horizontal flight, this reduces to P=NpQrω=QωP = N_p Q_r \omega= Q\omega, where QrQ_r is the torque of a single rotor and Q=NpQrQ = N_p Q_r is the total rotor torque.

Model-Based Analysis of Energy Efficiency

We first use the model of Sec. II to examine how vehicle mass affects energy-optimal steady horizontal flight in simulation. Under V˙x=0\dot{V}_x = 0, Vz=0V_z = 0, Θ˙=0\dot{\Theta} = 0, the metric for energy efficiency is the energy consumption per unit distance,

EPM=PVx=QωVx.(10)EPM = \frac{P}{V_x} = \frac{Q\omega}{V_x}. \tag*{(10)}

We consider mass-only variation of a fixed vehicle configuration, with its geometry, aerodynamic characteristics, and rotor configuration unchanged.

Figure 4 shows the simulated EPM per unit mass versus velocity curves for three vehicle masses. Each curve exhibits an energy-optimal velocity, which shifts to higher velocity as mass increases. The corresponding optimal operating points are summarized in Table 1. From these results, we report 3 important findings: the normalized optimal velocity Vx∗/mV_x^*/\sqrt{m}, optimal pitch angle Θ∗\Theta^*, and normalized minimum energy EPM∗/mEPM^*/m remain nearly constant across mass, suggesting the three scaling relationships

EPM per unit mass vs Horizontal Velocity under Different Vehicle Mass from Model Analysis

Figure 4. EPM per unit mass vs Horizontal Velocity under Different Vehicle Mass from Model Analysis

mm (kg)Vx∗V_x^* (m/s)Vx∗/mV_x^*/\sqrt{m}Θ∗\Theta^* (deg)EPM∗/mEPM^*/m (J/m/kg)
814.014.9515.386.19
1217.164.9515.386.20
1619.814.9515.376.20

Table 1. Simulated Energy-Optimal Operating Points

Vx∗∝m1/2,(11)V_x^* \propto m^{1/2}, \tag*{(11)}
EPM∗∝m,(12)EPM^* \propto m, \tag*{(12)}
Θ∗=const.(13)\Theta^* = \mathrm{const.} \tag*{(13)}

Note that the increase of Vx∗V_x^* with mm has also been observed in prior numerical studies [10], [14], but an explicit mass-scaling relationship was not established. A narrow range of energy-optimal Θ\Theta was similarly observed in [23], but its relationship with vehicle mass was not identified.

Derivation of Mass Scaling Laws for Energy-Optimal Flight

In this section, we prove the three mass-scaling relationships for energy-optimal flight in Eqs. (11)–(13). The derivation rests on a similarity property of the steady-state equations: at a fixed pitch angle Θ\Theta, the horizontal velocity, induced velocity, and rotor speed all scale with m\sqrt{m}, and hence their normalized quantities depend on Θ\Theta but not on mass. This similarity leads to P/m3/2=P′(Θ)P/m^{3/2} = P'(\Theta) and EPM/m=F(Θ)EPM/m = F(\Theta), where P′P' and FF are independent of mass. Consequently, the pitch angle minimizing EPM is independent of mass, and the scaling relationships for the optimal velocity and minimum EPM follow directly.

First, according to the force balance in Figure 2, we have

CBDVx2=Tsin⁡Θ,T=mgcos⁡Θ,(14)C_{BD}V_x^2 = T\sin\Theta,\qquad T = \frac{mg}{\cos\Theta}, \tag*{(14)}

which gives

Vx=mgCBDtan⁡Θ=mVx′(Θ),(15)V_x = \sqrt{\frac{mg}{C_{BD}}\tan\Theta} = \sqrt{m}V_x'(\Theta), \tag*{(15)}

where

Vx′(Θ)=gCBDtan⁡Θ,(16)V_x'(\Theta) = \sqrt{\frac{g}{C_{BD}}\tan\Theta}, \tag*{(16)}

is the desired mass-independent function of Θ\Theta.

Then, based on the Blade Element Momentum Theory in Eqn. (3) and (4), thrust TT and torque QQ can be re-written as

T=[BT1ω2+BT2Vx2cos⁡2Θ2−BT3ω(Vxsin⁡Θ+vi)](17)T = \left[B_{T1}\omega^2 + B_{T2}\frac{V_x^2\cos^2\Theta}{2} - B_{T3}\omega(V_x\sin\Theta+ v_i)\right] \tag*{(17)}

with constants

BT1=NpNbρa2∫R0Rr2c(r)θ(r) drB_{T1} = \frac{N_pN_b\rho a}{2}\int_{R_0}^{R}r^2c(r)\theta(r)\,dr
BT2=NpNbρa2∫R0Rc(r)θ(r) dr,BT3=NpNbρa2∫R0Rrc(r) dr(18)B_{T2} = \frac{N_pN_b\rho a}{2}\int_{R_0}^{R}c(r)\theta(r)\,dr,\qquad B_{T3} = \frac{N_pN_b\rho a}{2}\int_{R_0}^{R}rc(r)\,dr \tag*{(18)}

and

Q=CQ1ω2+CQ2Vx2cos⁡2Θ+CQ3(Vxsin⁡Θ+vi)ω−CQ4(Vxsin⁡Θ+vi)2(19)Q = C_{Q1}\omega^2 + C_{Q2}V_x^2\cos^2\Theta+ C_{Q3}(V_x\sin\Theta+ v_i)\omega- C_{Q4}(V_x\sin\Theta+ v_i)^2 \tag*{(19)}

with constants

CQ1=NpNbρcd2∫R0Rr3c(r) dr,CQ2=NpNbρcd4∫R0Rrc(r) dr,C_{Q1} = \frac{N_pN_b\rho c_d}{2}\int_{R_0}^{R}r^3c(r)\,dr,\qquad C_{Q2} = \frac{N_pN_b\rho c_d}{4}\int_{R_0}^{R}rc(r)\,dr,
CQ3=BT1,CQ4=BT3.(20)C_{Q3} = B_{T1},\qquad C_{Q4} = B_{T3}. \tag*{(20)}

Note that the thrust and torque of all propellers are equal under the steady-state flight, giving T=NpTrT = N_pT_r.

We now state the similarity property on which the three scaling laws rest.

Lemma 1 (Mass similarity of steady-state rotor solution). *Consider steady level flight of a multicopter at pitch angle

Θ∈(0,π/2)\Theta\in(0,\pi/2), with the body drag coefficient CBDC_{BD}, the rotor geometry and blade coefficients, and the air density held constant as the vehicle mass mm varies. Define the nondimensional induced velocity, tip-speed ratio and rotor thrust coefficient

vˉi=viVx,ωˉ=ωRVx,CT=TrρπR2Vx2.(21)\bar{v}_i = \frac{v_i}{V_x}, \qquad\bar{\omega} = \frac{\omega R}{V_x}, \qquad C_T = \frac{T_r}{\rho\pi R^2V_x^2}. \tag*{(21)}

Then CTC_T, vˉi\bar{v}_i and ωˉ\bar{\omega} are functions of Θ\Theta only and are independent of mm. Consequently, we have

vi=m vi′(Θ),vi′(Θ)=Vx′(Θ)vˉi(Θ)(22)v_i = \sqrt{m}\,v_i'(\Theta), \qquad v_i'(\Theta) = V_x'(\Theta)\bar{v}_i(\Theta) \tag*{(22)}
ω=m ω′(Θ),ω′(Θ)=Vx′(Θ)ωˉ(Θ)/R.(23)\omega= \sqrt{m}\,\omega'(\Theta), \qquad\omega'(\Theta) = V_x'(\Theta)\bar{\omega}(\Theta)/R. \tag*{(23)}

Proof. The force balance (14) gives Tr=mg/(Npcos⁡Θ)T_r = mg/(N_p\cos\Theta) and Vx2=(mg/CBD)tan⁡ΘV_x^2 = (mg/C_{BD})\tan\Theta. Hence their ratio

CT=mg/(Npcos⁡Θ)ρπR2(mg/CBD)tan⁡Θ=CBDNpρπR2sin⁡Θ,(24)C_T = \frac{mg/(N_p\cos\Theta)}{\rho\pi R^2(mg/C_{BD})\tan\Theta} = \frac{C_{BD}}{N_p\rho\pi R^2\sin\Theta}, \tag*{(24)}

in which the mass cancels identically. This is the single fact from which everything else follows.

Dividing the momentum-theory quartic (8) by Vx4V_x^4 and using (21) yield

vˉi2(vˉi2+2sin⁡Θ vˉi+1)=(CT2)2,(25)\bar{v}_i^2\left(\bar{v}_i^2 + 2\sin\Theta\,\bar{v}_i + 1\right) = \left(\frac{C_T}{2}\right)^2, \tag*{(25)}

whose coefficients only involve Θ\Theta (and constants) by (24). Hence the solution vˉi(Θ)\bar{v}_i(\Theta) only depends on Θ\Theta.

Dividing the thrust relation (17) by Vx2V_x^2, substituting T=mg/cos⁡ΘT = mg/\cos\Theta and Vx2=(mg/CBD)tan⁡ΘV_x^2 = (mg/C_{BD})\tan\Theta, and using (21) give

BT1R2ωˉ2−BT3Rωˉ(sin⁡Θ+vˉi)+BT22cos⁡2Θ=CBDsin⁡Θ,(26)\frac{B T_1}{R^2}\bar{\omega}^2 - \frac{B T_3}{R}\bar{\omega}(\sin\Theta+\bar{v}_i) + \frac{B T_2}{2}\cos^2\Theta= \frac{C_{BD}}{\sin\Theta}, \tag*{(26)}

whose coefficients again involve only Θ\Theta. Hence the solution is written as ωˉ(Θ)\bar{\omega}(\Theta). Since Vx=m Vx′(Θ)V_x = \sqrt{m}\,V_x'(\Theta) by (15), (22) and (23) follow directly from the definitions in (21). □

Next, the torque QQ from Eqn. (19) can be re-written by substituting the expressions for VxV_x in Eq. (15), viv_i in Eq. (22), and ω\omega in Eq. (23) as

Q=CQ1m ω′2(Θ)+CQ2m Vx′2(Θ)cos⁡2Θ+CQ3m(Vx′(Θ)sin⁡Θ+vi′(Θ))ω′(Θ)−CQ4m(Vx′(Θ)sin⁡Θ+vi′(Θ))2.(27)\begin{aligned} Q ={}& C_{Q1}m\,\omega'^{2}(\Theta) + C_{Q2}m\,V_x'^{2}(\Theta)\cos^2\Theta\\ &+ C_{Q3}m\left(V_x'(\Theta)\sin\Theta+v_i'(\Theta)\right)\omega'(\Theta) \\ &- C_{Q4}m\left(V_x'(\Theta)\sin\Theta+v_i'(\Theta)\right)^2. \tag*{(27)} \end{aligned}

Note that since Vx′V_x', vi′v_i', and ω′\omega' are all functions of only Θ\Theta, we can write QQ as

Q=mQ′(Θ).(28)Q = mQ'(\Theta). \tag*{(28)}

Eqn. (28) follows from the fact that every term of (19) is quadratic in the velocities, each of which scales as m\sqrt{m} by Lemma 1. It therefore holds irrespective of the numerical values of the blade constants.

Since the rotor torque scales as Q∝mQ \propto m by Eqn. (28) and the rotor speed as ω∝m\omega\propto\sqrt{m} by (23), the power required in steady horizontal flight satisfies

P(Θ;m)=Qω=m3/2P′(Θ),P′(Θ)=Q′(Θ)ω′(Θ).(29)P(\Theta;m) = Q\omega= m^{3/2}P'(\Theta), \qquad P'(\Theta) = Q'(\Theta)\omega'(\Theta). \tag*{(29)}

Combining (29) with Vx(Θ;m)=m Vx′(Θ)V_x(\Theta;m) = \sqrt{m}\,V_x'(\Theta) from (15), the mass-normalized energy per meter in (10) becomes

EPM(Θ;m)m=PmVx=m3/2P′(Θ)mmVx′(Θ)=P′(Θ)Vx′(Θ)≜F(Θ),(30)\frac{EPM(\Theta;m)}{m} = \frac{P}{mV_x} = \frac{m^{3/2}P'(\Theta)}{m\sqrt{m}V_x'(\Theta)} = \frac{P'(\Theta)}{V_x'(\Theta)} \triangleq F(\Theta), \tag*{(30)}

where FF is independent of mass for a fixed vehicle configuration. Therefore, the optimal pitch angle

Θ∗=arg⁡min⁡Θ EPM(Θ;m)=arg⁡min⁡Θ F(Θ),(31)\Theta^* = \underset{\Theta}{\arg\min}\ EPM(\Theta;m) = \underset{\Theta}{\arg\min}\ F(\Theta), \tag*{(31)}

is independent of mass, proving Eqn. (13). Evaluating Eqns. (15) and (30) at this common Θ∗\Theta^* gives

Vx∗=m Vx′(Θ∗)∝m1/2,(32)V_x^* = \sqrt{m}\,V_x'(\Theta^*) \propto m^{1/2}, \tag*{(32)}

and

EPM∗=mC,C=F(Θ∗)=Q′(Θ∗)ω′(Θ∗)Vx′(Θ∗),(33)EPM^* = mC, \qquad C = F(\Theta^*) = \frac{Q'(\Theta^*)\omega'(\Theta^*)}{V_x'(\Theta^*)}, \tag*{(33)}

proving Eqns. (11) and (12), respectively.

Remark 1 (Why the laws hold). Suppose the vehicle mass is multiplied by a factor kk. Every force in Figure 2 must then also be multiplied by kk. Every aerodynamic force, i.e. body drag, rotor thrust and torque, is quadratic in velocity, so multiplying every velocity by k\sqrt{k} multiplies every force by kk while leaving every angle unchanged: the pitch angle Θ\Theta, the inflow angle ϕ\phi of (4), and the blade angle of attack are all ratios of velocities and are therefore invariant. Heavier vehicles fly through a geometrically similar flow field simply faster. Mass sets the scale of the steady-state solution but not its shape. The three scaling laws are the three consequences of this similarity: angles are invariant, so the energy-optimal pitch angle is invariant; velocities scale as m\sqrt{m}, so the optimal cruise speed scales as m\sqrt{m}; and power is force times velocity and therefore scales as m3/2m^{3/2}, while energy per meter is power divided by velocity and therefore scales linearly in mass.

Experimental Validation

Outdoor flight tests were conducted on two multicopter platforms with different sizes, rotor counts, and propulsion systems at the Woodland-Davis Aeromodelers field. The octocopter is based on a DJI Spreading Wings S1000 equipped with a Pixhawk 2.1 flight controller and onboard GNSS, current, and voltage measurements, as shown in fig:1a. The quadrotor is a Hexsoon EDU-650 V2 equipped with a CubeOrange flight controller and similar onboard instrumentation.

The octocopter was tested at total masses of 7, 8, and 9 kg, spanning a 29% mass variation, while the quadrotor was tested at 5 masses from 3.58 to 5.17 kg, spanning 44%. Added payload was mounted compactly so that the projected area and overall aerodynamic shape remained approximately unchanged, consistent with the mass-only variation assumed in Sec. IV. Velocity set points were flown in both directions along a reciprocal straight-line course to mitigate directional wind effects through averaging. The campaigns are summarized in Table II. Power, horizontal velocity, and pitch angle were recorded onboard. Steady-flight segments were identified by requiring the ground speed to remain near its commanded set point for at least 3 s; all reported quantities were averaged over these segments. EPM was computed as ∫P dt/∫Vx dt\int P\,\mathrm{d}t / \int V_x\,\mathrm{d}t. Reciprocal measurements were averaged within each mass before cross-mass comparison. The pitch angle was obtained from the onboard flight-controller attitude estimate rather than inferred from the force-balance model.

PlatformMassesRange (kg)VariationSet pointsSteady legs
Quadrotor53.58–5.1744%655
Octorotor37.00–9.0029%6–739

Table II. Summary of experimental data.

Full-Range Validation of Similarity Relations

We first validate the scaling laws through the full-range similarity relationships. According to Eqs. (15) and (29), we have Vx/m=Vx′(Θ)V_x/\sqrt{m}=V_x'(\Theta) and P/m3/2=P′(Θ)P/m^{3/2}=P'(\Theta), which directly lead to EPM/m=F(Θ)EPM/m=F(\Theta) and hence the three optimal-flight scaling laws. We therefore examine whether Vx/mV_x/\sqrt{m} and P/m3/2P/m^{3/2} over Θ\Theta are invariant across different masses. We also validate the scaling laws using the experimentally identified optimal operating points in the next subsection. However, validation based on the optima requires the EPM minimum to be experimentally resolved at each mass, whereas the full-range approach uses the complete operating range including mass cases for which the optimum cannot be resolved.

Figure 5 and Figure 6 show the P/m3/2P/m^{3/2} and Vx/mV_x/\sqrt{m} versus Θ\Theta for all mass cases of each platform. The dashed curves show the corresponding model-based fits obtained from the steady-state dynamic and BEMT relations in Sec. II, with platform-specific parameters fitted to the experimental data. It is seen that both relationships show a strong collapse across mass. To quantify the agreement, the measurement data for each mass are interpolated to common values of Θ\Theta, and the cross-mass coefficients of variation (CoV) are computed. For quadrotor, the cross-mass CoV is 0.80–1.88% for P/m3/2P/m^{3/2} and 1.52–4.68% for Vx/mV_x/\sqrt{m}; for octorotor, the CoVs are 2.07–4.04% and 1.53–3.89%, respectively. The small CoVs indicate that P/m3/2P/m^{3/2} and Vx/mV_x/\sqrt{m} remain approximately mass-independent functions of Θ\Theta, directly validating the similarity relationships underlying the three optimal-flight scaling laws.

Plots of $P/m^{3/2}$ versus $\Theta$ for quadrotor and octorotor mass configurations

Figure 5. P/m3/2P/m^{3/2} versus Θ\Theta for all mass configurations of each platform.

Plots of $V_x/\sqrt{m}$ versus pitch angle for quadrotor and octorotor mass configurations

Figure 6. Vx/mV_x/\sqrt{m} versus Θ\Theta for all mass configurations of each platform.

Validation of Energy Optima

We next validate the scaling laws using the experimentally identified energy-optimal operating points. Figure 7 shows the mass-normalized EPM curves. For the heavier mass cases, the velocities required to bracket the EPM minimum exceeded the propulsion capability of the vehicle; consequently, the optimum can be identified for only 5 mass cases, as reported in Table III. Nevertheless, the full-range analysis in § 5-A provides validation across all mass configurations without requiring the EPM minimum to be resolved.

PlatformMass (kg)Θ∗\Theta^* (deg)Vx∗/mV_x^*/\sqrt{m}EPM∗/mEPM^*/m (J/m/kg)
Octorotor7.0016.345.068.31
Octorotor8.0016.194.958.17
Octorotor9.0015.504.928.01
Quadrotor3.5814.307.467.82
Quadrotor3.8314.827.677.84

Table III. Experimentally identified energy-optimal operating points.

Mass-normalized EPM versus horizontal velocity for the quadrotor and octorotor

Figure 7. Mass-normalized EPM versus horizontal velocity for the quadrotor (top) and octorotor (bottom). Curves show least-squares cubic fits to the measured set points.

For the octorotor, the identified Θ∗\Theta^* varies by only 0.84° across the 29% mass increase, while Vx∗/mV_x^*/\sqrt{m} and EPM∗/mEPM^*/m have coefficients of variation of 1.6% and 1.8%, respectively. Although EPM∗/mEPM^*/m shows a slight monotonic decrease with increasing mass from the graph, the variation remains small (1.8%) and represents only a minor deviation from the predicted mass invariance. For the two resolved quadrotor cases, Θ∗\Theta^* differs by 0.52°, while Vx∗/mV_x^*/\sqrt{m} and EPM∗/mEPM^*/m differ by 2.9% and 0.2%, respectively. These results are obtained from cubic fits to the measured EPM curves, providing consistent experimental validation for all three scaling laws on both platforms. To assess the sensitivity to curve fitting used to identify the optima, we additionally considered quadratic and quartic fits, as well as the physics-informed form EPM=a/Vx+b+cVx2EPM=a/V_x+b+cV_x^2, motivated by the standard induced, profile, and parasite power decomposition [11, 20]. Across the five resolved cases, the resulting Vx∗V_x^* values differ by only 2.7–4.9%, indicating that the choice of fitting function does not significantly affect the identified optima.

Discussion of Results

Significance of Findings

First, the linear scaling of the energy-optimal velocity V∗V^* with the square root of the total vehicle mass mm enables simple yet effective energy-efficient flight control protocol. Specifically, the energy-optimal velocity can be obtained as

Vx∗=mgCBDtan⁡Θ∗,(34)V_x^* = \sqrt{\frac{mg}{C_{BD}}\tan\Theta^*}, \tag*{(34)}

where the mass-invariant Θ∗\Theta^* only needs to be solved once, and then plugged in to obtain the Vx∗V_x^* under any mass mm. Alternatively, practitioners can manage to measure/identify Vx∗V_x^* under one specific mm through experiments [9, 12], and then scale with m\sqrt{m} to obtain Vx∗V_x^* under other mass conveniently. Consequently, the burden of repeated experimental measurements and/or online adaptation is significantly reduced.

Second, for multicopter energy consumption prediction and energy storage sizing, the found linear scaling of optimal EPMEPM with mass mm represents a highly useful advancement in practice. Specifically, the minimum energy consumption E∗E^* over a certain distance LL is

E∗=EPM∗L=CmL,(35)E^* = EPM^*L = CmL, \tag*{(35)}

which can be used to determine the minimum energy capacity needed for the specified mission distance. Conversely, given a fixed capacity of the onboard energy storage (e.g. battery) EE, the maximum range capability L∗L^* is

L∗=ECm.(36)L^* = \frac{E}{Cm}. \tag*{(36)}

The constant CC can be either computed theoretically based on Eqn. (33) or identified from experimental testing. These simple formulas enable quick yet accurate (theoretically proven) evaluation of the operation range or energy requirement.

Third, the closed-form energy model yields a lightweight cost metric for mission planning under payload variation, since each cruise segment has energy proportional to (mv+m)L(m_v+m)L when flown at the corresponding Vx∗V_x^*. This can be used to inform payload allocation and route selection without repeatedly solving an inner-loop velocity optimization:

min⁡{(mi,Li)}i=1NmE=∑iC(mv+mi)Li,(37)\min_{\{(m_i,L_i)\}_{i=1}^{N_m}} E = \sum_i C(m_v+m_i)L_i, \tag*{(37)}

where mvm_v is the vehicle mass. Again, it needs to be emphasized that all these advantages hinge on the condition that the multirotor is operated under the energy-optimal velocity which varies with the mass, but not any arbitrary velocity.

Finally, we remark that several commonly used mission-energy models in current literature directly assume a proportional relationship between vehicle mass and energy consumption rate EPMEPM under any horizontal velocity in the form of

EPM∝f(Vx)ma,(38)EPM \propto f(V_x)m^a, \tag*{(38)}

e.g. a=1a=1 in [6] and a=1.5a=1.5 in [17], and f(Vx)f(V_x) is a certain function of the velocity. Our results instead show that a simple linear-in-mass scaling holds only at the energy-optimal velocity, while the full EPMEPM–VxV_x curve does not scale uniformly with mm. The invariant flight state governing normalized efficiency is the pitch angle at optimum, Θ∗\Theta^*.

Impacts of Other Component Efficiency

As mentioned previously, other multicopter components, including the electric motor and the electronic speed controller (ESC), are not explicitly considered in the theoretical derivations of this work. Their efficiencies are treated as approximately constant over the flight regime of interest and therefore act as near-uniform multiplicative factors in the energy model, without altering the derived mass-scaling relationships. This assumption is supported by experimental results in the literature [8], which show that combined motor–ESC efficiency remains largely constant in the vicinity of the energy-optimal cruise velocity for a comparable multicopter platform, whereas the rotor aerodynamic efficiency (thrust per unit power) dominates the variation of the energy-per-meter (EPM) curve with respect to forward velocity.

Scope and Limitations

To clarify the scope of applicability of the derived scaling laws, we reiterate the key assumptions underlying the analysis. The scaling laws hold for mass-only variation of a fixed vehicle configuration in steady horizontal flight. Payload-induced center-of-mass shifts are assumed negligible, preserving the equal rotor thrust distribution assumed in Sec. II-B. Changes in airframe geometry or aerodynamic characteristics, e.g., the body drag coefficient, are also not considered in the present scope. Furthermore, transient maneuvers involving acceleration or attitude changes are excluded from the steady-state analysis.

Conclusion

This paper studies the scaling relationship between the energy-optimal flight behavior and multicopter mass under the steady-state horizontal flight. Several interesting and important findings have been made with both significant theoretical and practical implications. First, the energy-optimal pitch angle is found to be invariant to mass. This finding reflects the underlying mass-similarity property of steady horizontal flight: mass changes the scale of the steady-state solution, while the normalized aerodynamic state and the energy-optimal pitch angle remain unchanged. Second, the energy-optimal velocity is found to be proportional to the square root of the total mass. This finding enables convenient adaptation of flight control to mass/payload variation by only requiring a simple scaling of the velocity. Third, the optimal energy efficiency, measured by energy consumption per distance traveled, is found to be proportional to mass. This finding significantly simplifies the mission/task level energy-optimal planning problem, e.g. payload assignment for drone delivery, by enabling a simple linear problem formulation. The results not only advance the understanding of the fundamental multicopter energy dynamics, but also lay the foundation for energy-efficient operation and control optimization.

References

  1. [1]L. Bauersfeld and D. Scaramuzza, “Range, endurance, and optimal speed estimates for multicopters,” IEEE Robotics and Automation Letters, vol. 7, no. 2, pp. 2953–2960, 2022.arxiv.org/abs/2109.04741
  2. [2]T. Kirschstein, “Comparison of energy demands of drone-based and ground-based parcel delivery services,” Transportation Research Part D: Transport and Environment, vol. 78, p. 102209, 2020.DOI
  3. [3]C. C. Murray and R. Raj, “The multiple flying sidekicks traveling salesman problem: Parcel delivery with multiple drones,” Transportation Research Part C: Emerging Technologies, vol. 110, pp. 368–398, 2020.DOI
  4. [4]S. Poikonen and B. Golden, “Multi-visit drone routing problem,” Computers and Operations Research, vol. 113, p. 104802, 2020.DOI
  5. [5]J. K. Stolaroff, C. Samaras, E. R. O’Neill, A. Lubers, A. S. Mitchell, and D. Ceperley, “Energy use and life cycle greenhouse gas emissions of drones for commercial package delivery,” Nature Communications, vol. 9, no. 1, 2018.DOI
  6. [6]M. A. Figliozzi, “Lifecycle modeling and assessment of unmanned aerial vehicles (drones) CO2e emissions,” Transportation Research Part D: Transport and Environment, vol. 57, pp. 251–261, 2017.DOI
  7. [7]H. Li, Z. Zhan, and Z. Wang, “Energy-consumption model for rotary-wing drones,” Journal of Field Robotics, vol. 41, no. 6, pp. 1940–1959, 2024.DOI
  8. [8]N. Michel, P. Wei, Z. Kong, A. K. Sinha, and X. Lin, “Modeling and validation of electric multirotor unmanned aerial vehicle system energy dynamics,” eTransportation, vol. 12, p. 100173, 2022.DOI
  9. [9]C. Di Franco and G. Buttazzo, “Energy-aware coverage path planning of UAVs,” in 2015 IEEE International Conference on Autonomous Robot Systems and Competitions. IEEE, 2015, pp. 111–117.DOI
  10. [10]O. Doukkančić, B. Y. Kara, and T. Bektaş, “Minimizing energy and cost in range-limited drone deliveries with speed optimization,” Transportation Research Part C: Emerging Technologies, vol. 125, p. 102985, 2021.
  11. [11]Y. Zeng, J. Xu, and R. Zhang, “Energy minimization for wireless communication with rotary-wing UAV,” IEEE Transactions on Wireless Communications, vol. 18, no. 4, pp. 2329–2345, 2019.arxiv.org/abs/1804.02238
  12. [12]X. Wu and M. W. Mueller, “In-flight range optimization of multicopters using multivariable extremum seeking with adaptive step size,” in 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2020, pp. 1545–1550.arxiv.org/abs/2003.02907
  13. [13]N. Michel, P. Wei, Z. Kong, and X. Lin, “Energy-Optimal Unmanned Aerial Vehicles Motion Planning and Control Based on Integrated System Physical Dynamics,” Journal of Dynamic Systems, Measurement, and Control, vol. 145, no. 4, p. 041002, 01 2023.DOI
  14. [14]M.-h. Hwang, H.-R. Cha, and S. Y. Jung, “Practical endurance estimation for minimizing energy consumption of multirotor unmanned aerial vehicles,” Energies, vol. 11, no. 9, 2018.DOI
  15. [15]F. Tamke and U. Buscher, “The vehicle routing problem with drones and drone speed selection,” Computers & Operations Research, vol. 152, p. 106112, 2023.arxiv.org/abs/2111.13050
  16. [16]A. Patnaik, N. Michel, and X. Lin, Enhancing Multirotor Drone Efficiency: Exploring Minimum Energy Consumption of Forward Flight Under Varying Payload. AIAA SCITECH Forum, 2025. [Online]. Available: https://arc.aiaa.org/doi/abs/10.2514/6.2025-2187DOI
  17. [17]K. Dorling, J. Heinrichs, G. G. Messier, and S. Magierowski, “Vehicle routing problems for drone delivery,” IEEE Transactions on Systems, Man, and Cybernetics: Systems, vol. 47, no. 1, p. 70 – 85, 2017.DOI
  18. [18]R. D’Andrea, “Guest editorial can drones deliver?” IEEE Transactions on Automation Science and Engineering, vol. 11, no. 3, pp. 647–648, July 2014.
  19. [19]N. Michel, A. K. Sinha, Z. Kong, and X. Lin, “Multiphysical modeling of energy dynamics for multirotor unmanned aerial vehicles,” in 2019 International Conference on Unmanned Aircraft Systems (ICUAS). IEEE, 2019, pp. 738–747.DOI
  20. [20]J. G. Leishman, Principles of helicopter aerodynamics. Cambridge University Press, 2000.
  21. [21]J. D. Anderson, Fundamentals of aerodynamics, 6th ed., ser. McGraw-Hill series in aeronautical and aerospace engineering. New York, NY: McGraw Hill Education, 2017.
  22. [22]W. Johnson, Helicopter Theory. New York, NY: Dover Publications, 1994.
  23. [23]M. Perreault and K. Behdinan, “Delivery drone driving cycle,” IEEE Transactions on Vehicular Technology, vol. 70, no. 2, pp. 1146–1156, 2021.DOI

Paper details

vlnc:W_6mq2cba5
Version
v2 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Computer Science

Not recorded for this paper

View statement
Looking for related papers…
Contents