Variable-Tilt Omnidirectional Multirotor

• Variable-tilt omnidirectional multirotors are fully-actuated aerial platforms that use actively steerable rotors to independently control position and orientation in all six degrees of freedom.
• The design integrates rigid carbon fiber arms with tiltable brushless motors and a nonlinear control allocation model, enabling precise decoupling of translational and rotational dynamics.
• This platform supports applications in industrial inspection, aerial manipulation, cinematic motion, and agile navigation in constrained or dynamic environments.

A variable-tilt omnidirectional multirotor is a fully-actuated aerial robotic platform that employs rotors with actively steerable thrust directions. Unlike conventional underactuated multirotors, these vehicles are capable of independently controlling position and orientation in all six degrees of freedom. This architecture enables decoupling of translational and rotational dynamics, permitting advanced maneuvers, interactive tasks, and enhanced performance in complex or constrained environments.

1. Mechanical Architecture

Variable-tilt omnidirectional multirotors realize omnidirectional actuation by mounting each rotor at the end of a rigid arm with an integrated independent tilting mechanism. In the Voliro platform—a canonical example—six arms are radially distributed about the vehicle's main frame, each terminating in a tiltable rotor unit comprising a brushless DC motor and a dedicated tilting motor. The brushless tilting motors, typically 12 mm in diameter and equipped with high-precision Hall sensors and gearboxes, enable accurate angular control of up to 720°, limited only by cabling constraints. The thrust propeller is attached directly to the tilting module, while the carbon fiber tube arm remains fixed, maintaining global structural rigidity and preventing force transmission inefficiencies.

Actively steering the orientation of the thrust vector allows each rotor to not only contribute to vertical lift but also to lateral force and torque generation. The geometric configuration ensures each rotor's contribution covers the required force/moment space under arbitrary platform attitudes, a critical design consideration to avoid dead zones or force counteraction.

2. Actuation Model and Control Allocation

The fundamental challenge for control stems from the overactuation: for a vehicle with $n$ tiltable rotors, there are $2n$ independent inputs (rotor speed squared $n_i^2$ and tilt angle $\alpha_i$ for each rotor) and $6$ output degrees of freedom (forces and moments in $\mathbb{R}^6$). The mapping from actuator commands to vehicle wrench is nonlinear due to the coupling between rotor speed and tilt angle. For Voliro, the relation is

$$\left[\begin{array}{c} \mathbf{F} \ \mathbf{M} \end{array}\right]=A(\boldsymbol{\alpha})\mathbf{N}$$

with $$\mathbf{N} = [n_1^2,\, n_2^2,\, \ldots,\, n_6^2]^\top$$ and allocation matrix $A$ depending nonlinearly on the tilt angles $\boldsymbol{\alpha}$. The inversion of this mapping is nontrivial due to these nonlinearities and the actuation redundancy.

To resolve this, the force for each rotor is projected onto “vertical” and “lateral” axes: $$F_{v,i} = \mu\,n_i^2\,\cos(\alpha_i), \quad F_{l,i} = \mu\,n_i^2\,\sin(\alpha_i)$$ This allows recasting the allocation as a linear problem in this augmented $[F_{v,1},F_{l,1},\ldots,F_{v,6},F_{l,6}]$ space, where a static allocation matrix $A_{\text{static}}$ maps the desired wrench to the stacked forces. The Moore–Penrose pseudoinverse of $A_{\text{static}}$ is used for fast onboard computation, and individual actuator commands are recovered by

$$n_i^2 = \frac{1}{\mu}\sqrt{F_{v,i}^2 + F_{l,i}^2}, \qquad \alpha_i = \arctan2(F_{l,i}, F_{v,i})$$

This approach enables real-time computation at several hundred Hz and decouples position and attitude control.

3. Feedback and Reference Tracking

Position control is performed by a PID controller with gravity and feedforward acceleration compensation: $${}_{(B)}F_{\text{des}} = R_{IB}^\top[ k_{p,p}(\mathbf{p}_{\text{des}}-\hat{\mathbf{p}}) + k_{d,p}(\dot{\mathbf{p}}_{\text{des}}-\hat{\dot{\mathbf{p}}}) + k_{i,p}\int(\mathbf{p}_{\text{des}}-\hat{\mathbf{p}})dt + mg + \ddot{\mathbf{p}}_{\text{des}}]$$ Attitude regulation employs a cascade structure: outer-loop quaternion error tracking,

$$q_{\text{err}} = q_{\text{des,IB}}\otimes \hat{q}_{IB}^*$$

with axis-angle mapped desired body rates,

$$\boldsymbol{\omega}_{\text{des}} = k_q \cdot \text{sign}(q_{w,\text{err}})\cdot \mathbf{q}_{v,\text{err}}$$

passed to an inner rate controller for moment generation.

4. Design Challenges: Mechanical and Control

Integrating a high-precision tilting mechanism in lightweight arms required embedding the brushless motor into the carbon tube, balancing structural integrity with weight minimization and avoiding deformation of arms that would undermine the allocation model. Rotor placement on a circular array about the center of gravity ensures effective force contribution under arbitrary tilt angles.

From the control perspective, the actuation redundancy introduces a nullspace of solutions: not all required wrenches map uniquely to an actuator configuration. Variations in tilting motor dynamics pose further challenges—tilt actuators are substantially slower than thrust motors, limiting responsiveness during aggressive maneuvers. Under certain configurations (e.g., 90° pitch), some rotors can become ineffective for gravity compensation, requiring adaptive allocation strategies.

5. Experimental Demonstration

Extensive validation comprised simulated and real flight tests:

• Simulation was conducted in Gazebo (RotorS plugin), capturing full sensor, motor, and tilt dynamics.
• Free Flight: Maneuvers included transitions from upright to inverted orientation (rotating about body–y axis while maintaining control) and sustained 50° pitched flight with horizontal translation. The system tracked position and attitude accurately, with actuator commands reflecting allocation to regions of varying controllability.
• Wall Contact Interaction: A three-sphere compliance module was attached, and the platform executed vertical flight, approach, and contact-driven “driving” along a wall. Even with slower tilt dynamics, the controller maintained stable force application and controlled motion along the vertical surface, with only minor coupling in the most demanding configurations.

6. Application Domains

The variable–tilt omnidirectional multirotor's capability for decoupled force and torque generation underpins several key application areas:

• Industrial Inspection: Allows contact-sustained inspection over curved and irregular geometries (e.g., bridge or facade inspection) while maintaining continuous sensor alignment.
• Aerial Manipulation: Full pose control supports maintenance and forceful environmental interaction, such as wall cleaning, tool-based inspection, or sample collection.
• Cinematic Motion: Uninterrupted, arbitrary camera trajectories—decoupling orientation from translation—for smooth and complex shot generation in film production.
• Agile Navigation and Surveillance: Omnidirectional translation and attitude adjustment enable agile navigation in cluttered, three-dimensional environments.

7. Performance Limitations and Extensions

The main performance bottleneck remains the slower dynamics of tilting motors relative to thrust motors, which can induce transient errors and potential actuator saturation in extreme configurations. Allocation schemes must dynamically adapt to underactuated subspaces when some rotors are unavailable for gravity compensation. The demonstrated control algorithms and mechanical solutions, however, already enable complex maneuvers and interaction tasks unobtainable by standard multi-copters.

In summary, the variable–tilt omnidirectional multirotor exemplifies an effective approach to achieving full six–degree–of–freedom aerial control in a compact platform. This capability arises from the integration of independent tilting actuation at each rotor, a nonlinear control allocation framework, and cascaded position/attitude regulation. The architecture significantly broadens the spectrum of agile flight and forceful interaction tasks feasible in real-world scenarios, as evidenced by validated experiments on prototypes such as Voliro (Kamel et al., 2018).