---
title: Distributed Nonlinear Dynamic Inversion (DNDI)
url: https://www.emergentmind.com/topics/distributed-nonlinear-dynamic-inversion-dndi
type: topic
---

# Distributed Nonlinear Dynamic Inversion (DNDI)

Distributed Nonlinear Dynamic Inversion (DNDI) denotes a class of nonlinear control designs in which local nonlinear dynamics are inverted at the agent level while coordination objectives are imposed through distributed interaction laws over a communication graph. In the material presently available, the clearest explicit formulation appears in cooperative missile guidance, where DNDI is defined as the combination of nonlinear dynamic inversion or input–output linearization for each missile and a distributed consensus-tracking protocol that uses local and neighbor information to enforce coordinated terminal behavior [2509.18022]. Closely related work extends the surrounding landscape in three directions: robust incremental inversion for aerial robots [2501.07223], data-driven online inversion of identified nonlinear predictors [1407.1069], and distributed nonlinear control frameworks that are inverse-like but not based on exact cancellation, such as separable control contraction metrics [1810.04794] and distributed inverse dynamics for floating-base legged systems [2412.09816]. Taken together, these works place DNDI at the intersection of nonlinear inversion, distributed coordination, and practical constraint handling.

## 1. Definition, scope, and conceptual position

In the cooperative missile-guidance formulation, DNDI extends nonlinear dynamic inversion to a distributed multi-agent setting: each missile inverts its own nonlinear guidance dynamics, while the desired closed-loop behavior is expressed not as purely local tracking but as distributed consensus tracking with respect to neighbors and a virtual leader [2509.18022]. In that formulation, DNDI is therefore the synthesis of two ingredients: input–output linearization at the agent level and graph-based leader-following coordination at the network level.

This usage is narrower than a generic phrase such as “distributed inversion-based control.” The missile-guidance paper is explicitly about multiple follower missiles attacking a stationary ground target, with cooperative variables chosen as time-to-go and line-of-sight (LOS) rate [2509.18022]. By contrast, the other supplied papers do not formulate DNDI under that name. The robust aerial-robotics work develops a cascaded single-agent INDI architecture with robust outer-loop synthesis rather than distributed coordination [2501.07223]. The nonlinear inversion control paper presents a centralized, data-driven SISO inversion method for unknown discrete-time plants [1407.1069]. The separable-CCM paper provides a distributed nonlinear tracking framework without exact plant inversion [1810.04794]. The quadruped paper uses distributed inverse dynamics and constrained contact-force allocation rather than graph-based multi-agent coordination [2412.09816].

Accordingly, DNDI should be understood in two layers. In the strict sense represented directly in the literature excerpt, it is a distributed input–output linearization method with consensus-tracking error dynamics [2509.18022]. In a broader research sense, the surrounding papers suggest a family resemblance among methods that separate local nonlinear inversion or inverse-dynamics realization from higher-level distributed objectives, but that broader interpretation is an inference rather than a formal shared definition.

## 2. Canonical mathematical structure

The generic nonlinear agent used in the DNDI missile-guidance formulation is written as
\[
\dot X_i = f(X_i)+g(X_i)U_i,
\qquad
Y_i = h(X_i).
\]
Differentiating the output yields
\[
\dot Y_i
=
\left[\frac{\partial h}{\partial X_i}\right]\dot X_i
=
\left[\frac{\partial h}{\partial X_i}\right]\big(f(X_i)+g(X_i)U_i\big)
=
f_Y(X_i)+g_Y(X_i)U_i,
\]
with
\[
f_Y(X_i)=\left[\frac{\partial h}{\partial X_i}\right]f(X_i),
\qquad
g_Y(X_i)=\left[\frac{\partial h}{\partial X_i}\right]g(X_i).
\]
This is the standard dynamic-inversion step at the output level [2509.18022].

The distributed extension is introduced through the consensus-tracking error. For scalar output,
\[
e_i = \sum_{j\in N_i} a_{ij}(y_i-y_j)+\beta_i(y_i-y_L),
\]
and for vector output \(Y_i\in\mathbb{R}^p\),
\[
E_i = (\bar d_i+\bar\beta_i)Y_i-\bar a_i \mathbf Y-\bar\beta_i Y_L,
\]
equivalently
\[
E_i = \sum_{j\in N_i} a_{ij}(Y_i-Y_j) + \beta_i(Y_i-Y_L).
\]
The desired error dynamics are chosen as
\[
\dot E_i + K_i E_i = 0.
\]
Substituting the output dynamics and solving algebraically for the control gives the distributed inversion law
\[
U_i = (g_Y(X_i))^{-1} \Bigg[ -f_Y(X_i) + (\bar d_i+\bar\beta_i)^{-1} \Big( \bar a_i \dot{\mathbf Y} +\bar\beta_i \dot Y_L - K_i\big( (\bar d_i+\bar\beta_i)Y_i-\bar a_i\mathbf Y-\bar\beta_i Y_L \big) \Big) \Bigg].
\]
This equation is the central DNDI law in the missile-guidance paper [2509.18022].

The communication model is a weighted undirected connected graph
\[
G=\{V,E\},
\]
with adjacency matrix \(A=[a_{ij}]\), degree matrix
\[
D=\mathrm{diag}\{d_1,\dots,d_N\},\qquad d_i=\sum_{j\in N_i} a_{ij},
\]
and Laplacian
\[
L=D-A.
\]
Leader access is encoded by \(\beta_i\) [2509.18022]. The resulting architecture is distributed in the sense that each agent uses its own output, neighboring outputs, and, if applicable, leader information. The missile-guidance formulation further requires neighbor output derivatives \(\dot{\mathbf Y}\), or quantities from which they can be obtained [2509.18022].

This structure differs sharply from centralized nonlinear inversion of a single plant. In the data-driven nonlinear inversion control method, for example, the controller is a centralized one-step right-inversion of an identified predictor
\[
\hat{y}_{t+1}=f(\boldsymbol q_t,u_t),
\]
implemented by
\[
u_t^{nl}=\arg\min_{\mathfrak{u}\in U}J(\mathfrak{u}),
\]
with the inversion performed over a scalar constrained optimization rather than over a graph-coupled consensus law [1407.1069]. That contrast clarifies what is specifically “distributed” in DNDI: not merely local inversion, but inversion embedded inside networked cooperative error dynamics.

## 3. Missile-guidance formulation and cooperative objectives

The most explicit DNDI application in the supplied material concerns a salvo of follower missiles engaging a stationary ground target through a virtual-leader architecture [2509.18022]. The purpose is simultaneous target interception rather than isolated individual interception. The paper states two control objectives: the follower missiles are required to track a virtual time-to-go profile, and the tracking of zero LOS rate assures interception [2509.18022].

For the \(i\)-th missile, the planar motion model is
\[
\dot{V}_i = a_{t_i},
\qquad
\dot{\gamma}_i = \frac{a_{n_i}}{V_i},
\qquad
\dot{x}_i = V_i \cos \gamma_i,
\qquad
\dot{z}_i = V_i \sin \gamma_i.
\]
With \(r_i\) the relative distance, \(\lambda_i\) the LOS angle, and \(\phi_i=\gamma_i-\lambda_i\), the relative kinematics are
\[
\dot{r}_i = -V_i \cos \phi_i,
\qquad
\dot{\lambda}_i = -\frac{V_i \sin \phi_i}{r_i}.
\]
Differentiation yields
\[
\ddot{r}_i = r_i \dot{\lambda}_i^2 + u_{i_1},
\qquad
\ddot{\lambda}_i = -\frac{2\dot{r}_i \dot{\lambda}_i}{r_i} - \frac{u_{i_2}}{r_i},
\]
with transformed inputs
\[
\begin{bmatrix} u_{i_1}\\ u_{i_2} \end{bmatrix}
=
\begin{bmatrix}
\cos\phi_i & -\sin\phi_i\\
\sin\phi_i & \cos\phi_i
\end{bmatrix}
\begin{bmatrix} a_{t_i}\\ a_{n_i} \end{bmatrix}.
\]
This orthogonal transformation maps tangential and normal acceleration commands into range-aligned and LOS-aligned virtual inputs [2509.18022].

The state variables are
\[
x_{i_1}=r_i,\qquad x_{i_2}=\dot r_i,\qquad x_{i_3}=\lambda_i,\qquad x_{i_4}=\dot\lambda_i,
\]
and the chosen outputs are
\[
Y_i= \begin{bmatrix} t_{go_i}\\ \dot{\lambda}_i \end{bmatrix},
\qquad
t_{go_i} = -\frac{x_{i_1}}{x_{i_2}}.
\]
Their first derivatives are
\[
\dot t_{go_i} = -1+\frac{x_{i_1}^2x_{i_4}^2}{x_{i_2}^2} -\frac{x_{i_1}}{x_{i_2}^2}u_{i_1},
\]
and
\[
\dot{(\dot\lambda_i)}=\ddot\lambda_i = -\frac{2x_{i_2}x_{i_4}}{x_{i_1}}-\frac{u_{i_2}}{x_{i_1}}.
\]
Hence the output vector has relative degree \((1,1)\) with respect to
\[
U_i=[u_{i_1},u_{i_2}]^T,
\]
and
\[
g_Y(X_i)=
\begin{bmatrix}
-\dfrac{x_{i_1}}{x_{i_2}^2} & 0\\[8pt]
0 & -\dfrac{1}{x_{i_1}}
\end{bmatrix}.
\]
The inversion matrix is diagonal and nonsingular provided \(x_{i_1}=r_i\neq 0\) and \(x_{i_2}=\dot r_i\neq 0\) [2509.18022].

The virtual leader output is defined as
\[
Y_L=
\begin{bmatrix}
\delta t+t_{go_0}\\
0
\end{bmatrix},
\qquad
\delta<0,
\]
with \(t_{go_0}\) described as an average initial or final interception time reference [2509.18022]. Distributed coordination is then imposed on the vector
\[
Y_i=
\begin{bmatrix}
t_{go_i}\\
\dot\lambda_i
\end{bmatrix},
\]
so that time-to-go consensus implies simultaneous impact and LOS-rate regulation implies a collision course.

The same paper states that all missiles hit the target simultaneously in simulation, that \(t_{go_i}\) reaches consensus and tracks the leader time-to-go profile, that relative ranges converge to zero simultaneously, and that LOS rates converge to zero [2509.18022]. At the same time, its validation remains qualitative: no exact impact-time errors, miss distances, convergence times, peak acceleration commands, or robustness margins are reported in the provided text [2509.18022]. The paper also does not provide a new Lyapunov proof for the missile-specific closed loop; the exponential decay of \(E_i\) follows directly only from the imposed first-order error model, contingent on valid inversion and persistence in the nonsingular operating region [2509.18022].

## 4. Local inversion mechanisms, robustness, and realizability

The broader DNDI problem is not only one of distributed coordination; it is also one of ensuring that each agent can realize the commanded virtual controls robustly under modeling error, disturbances, and sensing limitations. The aerial-robotics INDI work is especially relevant on this point [2501.07223].

That paper contrasts classical nonlinear dynamic inversion for
\[
\dot x = f(x) + g(x)u
\]
with incremental nonlinear dynamic inversion. Classical NDI depends on explicit cancellation or inversion of the nonlinear model, so performance can degrade when \(f(x)\), \(g(x)\), actuator effectiveness, or disturbances are uncertain [2501.07223]. INDI instead uses a first-order expansion around the previous sample \((x_0,u_0)\),
\[
\dot x = \dot x_0 + \left.\frac{\partial [f(x)+g(x)u_0]}{\partial x}\right|_{x=x_0}(x-x_0) + g(x_0)(u-u_0),
\]
neglects the state increment term at high sampling rate, and obtains
\[
u = u_0 + g(x_0)^\dagger \bigl(v-\dot x_0\bigr).
\]
The control depends on control effectiveness and measured state-derivative data, but not directly on the uncertain drift term \(f(x)\) [2501.07223]. This measured-increment structure is the source of INDI robustness to model mismatch.

The same paper develops a full cascaded architecture for a multirotor drone. In the translational loop, a position controller generates commanded translational accelerations \(v_{ac}\), and an outer INDI law converts these into commanded thrust magnitude and Euler angles. In the rotational loop, an attitude controller generates commanded angular accelerations \(v_i\), and an inner INDI law converts these into motor angular-velocity commands [2501.07223]. The relevance to DNDI lies in the modular decomposition: local low-level inversion can remain onboard each vehicle, while an outermost command generator could, in principle, be replaced by distributed coordination laws [2501.07223].

For the outer guidance loop, the incremental translational relation is
\[
f_{Bc}b_{zc} = m(v_{ac}-a_f) + f_{Bf}b_{zf},
\]
and for the inner stabilization loop the commanded motor angular velocity is
\[
\omega_c = \omega_f + (G_1+G_2)^\dagger \left( v_c - \dot\Omega_f + G_2L(\omega_c-\omega_f) - \frac{\dot T_c}{m} \right).
\]
Disturbance compensation is implicit rather than observer-based: disturbance forces and torques are reconstructed through measured-filtered accelerations and rates [2501.07223]. The paper stresses that the same second-order filter must be used consistently for all incremented and measured signals to keep timing aligned [2501.07223].

The robust outer-loop augmentation uses low-order structured \(\mathcal H_\infty\) synthesis. The generalized plant is designed so that the controller \(K(s)\) minimizes the closed-loop induced norm from
\[
w=[r,d,n]
\]
to weighted performance outputs
\[
z=[z_1,z_2,z_3],
\]
with inequalities such as
\[
\|W_e(s)S(s)\|_\infty \le \gamma,
\qquad
\|W_a(s)S_{a_i}(s)\|_\infty \le \gamma,
\qquad
\|W_uK(s)S(s)\|_\infty \le \gamma,
\qquad
\|W_nS_n(s)\|_\infty \le \gamma.
\]
The structured low-order controller is synthesized via nonsmooth optimization using MATLAB’s `systune` [2501.07223].

Quantitatively, the reduced-order INDI/\(\mathcal H_\infty\) controller improves disturbance attenuation by nearly \(50\%\) relative to INDI/PD in the attitude loop, while the full-order version improves it by more than \(70\%\); in the position loop under force disturbances of amplitude between \(1.5\) N and \(6\) N, the reduced-order INDI/\(\mathcal H_\infty\) controller improves disturbance attenuation by about \(70\%\) [2501.07223]. The abstract and conclusion state that disturbance rejection improves by more than \(50\%\) overall for both rotational and translational dynamics [2501.07223]. This is not distributed control, but it identifies a local-agent robustness problem that any practical DNDI system must solve.

The data-driven inversion paper adds a different realizability perspective. There the plant is an unknown nonlinear discrete-time SISO system
\[
y_{t+1}=g\left(\boldsymbol{y}_{t},\boldsymbol{u}_{t},\boldsymbol{\xi}_{t}\right),
\]
approximated by a polynomial predictor
\[
\hat{y}_{t+1}=f\left(\boldsymbol{q}_{t},u_{t}\right)
=\sum_{i=1}^{N}\alpha_{i}\phi_{i}\left(\boldsymbol{q}_{t},u_{t}\right),
\]
with online inversion
\[
u_{t}^{nl}=\arg\min_{\mathfrak{u}\in U}J\left(\mathfrak{u}\right).
\]
Because \(J(\mathfrak u)\) is polynomial in a scalar decision variable, the candidate minimizers are the real roots of \(dJ/d\mathfrak u\) in the admissible set together with the saturation endpoints [1407.1069]. This is not DNDI, but it shows a distinct route to nonlinear inversion when analytical models are unavailable: identify a local predictor from data, then perform constrained right-inversion online. A plausible implication is that future DNDI systems could combine distributed coordination with local identification-and-inversion primitives rather than exact first-principles models.

## 5. Related distributed nonlinear control frameworks

DNDI exists alongside distributed nonlinear control frameworks that achieve trajectory tracking or stabilization without exact input–output cancellation. The clearest example in the supplied material is the separable-CCM approach [1810.04794].

That paper considers networked nonlinear systems on a physical interaction graph \(\mathscr G_p\) and a communication graph \(\mathscr G_c\), with local subsystem dynamics
\[
\dot{x}_i(t)=f_i(x_i(t),\breve{x}_i(t))+b_i(x_i(t),\breve{x}_i(t))u_i(t),
\]
and stacked form
\[
\dot{x}(t)=f(x(t))+B(x(t))u(t).
\]
The goal is universal exponential stabilizability of any forward-complete target trajectory \((x^\star,u^\star)\), under controllers that are \(\mathscr G_c\)-admissable in the sense that local input \(u_i\) depends only on local and communicated neighbor states and targets [1810.04794].

The core condition is the existence of a dual metric \(W(x)\) and structured matrix function \(Y(x)\) satisfying the pointwise LMI
\[
-\dot{W}+AW+WA^T+BY+(BY)^T+2\lambda W\prec0.
\]
With \(K(x)=Y(x)W^{-1}(x)\), the differential controller is
\[
\delta_u=K(x)\delta_x,
\]
and the actual nonlinear feedback is reconstructed along a minimizing geodesic \(\gamma\) by
\[
u(t)=u^\star(t)+\int_0^1 K(\gamma(t,s))\gamma_s(t,s)\,ds.
\]
Distribution is enforced structurally: \(W\) is block diagonal with local dependence, and \(Y\) respects the communication sparsity pattern [1810.04794]. The method is therefore distributed, nonlinear, and tracking-oriented, but it is not exact dynamic inversion.

Its significance for DNDI is comparative. The paper is explicit that the framework is useful for systems that are not feedback linearizable and even for a network with over \(1000\) states whose linearization is uncontrollable [1810.04794]. This stands in contrast to exact DNDI formulations that rely on the invertibility of \(g_Y(X_i)\) and well-defined relative degree. The CCM result therefore functions as an alternative design doctrine: instead of canceling nonlinearities, impose contraction through convex synthesis and geodesic reconstruction.

The quadruped paper offers another adjacent architecture. The full floating-base robot dynamics are
\[
\mathbf{M}(\mathbf{q})\ddot{\mathbf{q}} + \boldsymbol{\eta}(\mathbf{q},\dot{\mathbf{q}})
= \mathbf{S}^T\boldsymbol{\tau} + \mathbf{J}_c^T\mathbf{F}_c,
\]
and the controller first computes a virtual generalized force as if the \(18\)-DoF system were fully actuated:
\[
\boldsymbol{\tau}_f = \hat{\mathbf{M}}(\mathbf{q})\ddot{\mathbf{q}}_{cmd} + \hat{\boldsymbol{\eta}}(\mathbf{q},\dot{\mathbf{q}}).
\]
Under perfect modeling,
\[
\ddot{\mathbf{q}} = \ddot{\mathbf{q}}_{cmd},
\]
which is a computed-torque or dynamic-inversion relation [2412.09816]. The generalized force is then partitioned into base and joint components, with the base wrench realized through stance-contact forces by solving
\[
\mathbf{J}_{ab}^T\mathbf{F}_c = \boldsymbol{\tau}_b
\]
subject to exact friction-cone and unilateral constraints, and the final joint torque chosen as
\[
\boldsymbol{\tau} = -\mathbf{J}_{aa}^T\mathbf{F}_c^* + \mathbf{N}_{a,b}\boldsymbol{\tau}_j.
\]
The orthogonality condition
\[
(\mathbf{J}_{a,b})^\dagger \mathbf{N}_{a,b} = \mathbf{0}
\]
ensures that the projected joint-tracking torque does not alter the effective base wrench [2412.09816].

This is not DNDI in the missile-guidance sense, because the “distribution” is across contact channels and task subspaces rather than across communicating agents. Yet it illustrates how inversion-based nonlinear control becomes constrained force allocation in underactuated systems. The paper reports that its geometric projected-gradient solver is about \(2.47\times\) faster than qpOASES, with about \(31\%\) lower residual and about \(99\%\) lower constraint violation, while the overall controller reduces foot slip, improves orientation tracking, and uses about \(5\%\) less power than a QP-based balance controller in the reported tests [2412.09816]. For DNDI, the broader lesson is that exact inversion alone is often insufficient; realization constraints determine the actual control architecture.

## 6. Limitations, misconceptions, and likely research directions

A common misconception is that DNDI is simply distributed consensus appended to any nonlinear controller. The missile-guidance formulation is more specific: it requires a well-defined input–output map
\[
\dot Y_i=f_Y(X_i)+g_Y(X_i)U_i
\]
with invertible \(g_Y(X_i)\), and it embeds graph-based consensus directly at the output-derivative level [2509.18022]. Another misconception is that any distributed nonlinear controller is a DNDI controller. The separable-CCM method is distributed and nonlinear but avoids exact inversion altogether [1810.04794]; the quadruped method uses inverse dynamics but not multi-agent consensus [2412.09816]. Terminological precision is therefore important.

The missile-guidance DNDI paper also has explicit limitations. It does not treat impact-angle control, target assignment among multiple targets, inter-missile spacing or collision constraints, actuator-constrained optimality, or finite-time proofs [2509.18022]. It does not explicitly model communication delay, packet loss, asynchronous update, bandwidth limits, quantization, or exchanged-data noise [2509.18022]. It does not provide a missile-specific Lyapunov proof, a boundedness proof for all internal states, or a singularity-avoidance argument for \(r_i\to 0\) or \(\dot r_i\to 0\) [2509.18022]. These omissions are not peripheral: they identify the gap between a clean DNDI derivation and a fully fieldable distributed guidance law.

The aerial-robotics INDI work exposes a different set of limitations that carry over naturally to DNDI-style systems. Incremental inversion depends on high controller update rate, negligible state change over one sample, slowly varying nonlinear terms relative to the sample period, and sufficiently accurate local control-effectiveness estimates [2501.07223]. Differentiation of noisy gyro signals requires careful filtering, which introduces phase lag [2501.07223]. The approach also relies on reasonably accurate local matrices such as \(g(x_0)\), \(G_1\), and \(G_2\); severe mismatch can degrade inversion [2501.07223]. These are local-agent constraints, but distributed coordination cannot compensate for them.

The data-driven inversion paper highlights a further issue: scalability. Its polynomial root-finding inversion is computationally attractive because it is one-dimensional and SISO [1407.1069]. Extending that exact mechanism to MIMO or networked settings would no longer reduce to a scalar polynomial stationary condition [1407.1069]. This suggests that data-driven DNDI would require new decomposition or approximation strategies rather than a direct reuse of the SISO method.

A plausible research trajectory, suggested by the supplied materials but not proved in any single paper, is a layered architecture in which distributed coordination generates local virtual commands, while each agent realizes those commands through a robust local inversion layer. The robust INDI cascaded architecture already exhibits the modular separation needed for such a design [2501.07223]. The missile-guidance DNDI law provides the graph-coupled cooperative layer [2509.18022]. The data-driven inversion paper suggests how local inversion might be learned when accurate models are unavailable [1407.1069]. The CCM framework suggests a fallback when exact inversion is structurally impossible or too brittle [1810.04794]. The quadruped work suggests how realization constraints can be absorbed into optimization and projection layers rather than analytic inversion alone [2412.09816].

In that sense, DNDI is best regarded not as a single finished methodology but as a design pattern whose strictest current form is distributed input–output linearization with consensus tracking, and whose practical development depends on robust local inversion, communication-aware coordination, and feasibility-preserving realization layers.

Source: https://www.emergentmind.com/topics/distributed-nonlinear-dynamic-inversion-dndi