Drones Can Now Safely Operate Near People Thanks to New Impact Limits

Researchers are addressing a critical gap in the safe deployment of indoor micro-aerial vehicles (MAVs) by developing a method to translate impact testing into operational safety limits. Aziz Mohamed Mili, Louis Catar, and Paul Gérard, from Lab INIT Robots at the Department of Mechanical Engineering, Ecole de technologie superieure, alongside Ilyass Tabiai and David St-Onge et al., present an open toolchain that converts benchtop impact tests into deployable safety governors for drones. This work is significant because it provides a practical, data-driven approach to tuning motion limits based on measured impact risk, something currently lacking in the field. By offering shareable datasets, repeatable processes, and a ROS2 node for online enforcement, the team facilitates a pathway to certify indoor MAV operations near humans and preserve task performance while adhering to specified force constraints.

Drone impact data informs real-time safety governor deployment

Researchers have developed a comprehensive toolchain that directly translates drone impact test data into deployable safety governors for micro-aerial vehicles. Addressing a critical gap in indoor drone operation, this work provides a practical method for tuning motion limits based on measured impact risk, a capability previously lacking for practitioners.
The study details a replicable impact rig and protocol used to capture precise force-time profiles across various drone classes and contact surfaces, generating a robust dataset for analysis. Data-driven models were then created to map pre-impact speed to both impulse and contact duration, allowing for the direct computation of safe speed bounds for a specified force limit.

This innovative system goes beyond simply identifying impact forces; it delivers a functional solution for real-time safety enforcement. Scripts and a ROS2 node were released to implement these speed bounds online, logging compliance and accommodating facility-specific safety policies. Validation on commercial off-the-shelf quadrotors, including the Bamboo Cognifly, Carbon Cognifly, DJI Avata, and Flywoo Flylens, alongside representative indoor assets, demonstrated that the derived governors maintain task performance while adhering to stringent force constraints.

The core contribution of this research is a practical bridge connecting measured impact data to runtime control limits, offering shareable datasets, code, and a repeatable process. This enables teams to confidently certify indoor MAV operations in close proximity to humans, moving beyond reliance on finite-element models and towards empirically validated safety measures.

By quantifying the relationship between impact orientation, structural compliance, and peak force, the study highlights the importance of rebound as a critical safety parameter. This work establishes a foundation for certifiable, facility-specific speed governance, paving the way for wider adoption of indoor drone technology.

Drone impact testing via linear catapult and triaxial load cell measurement

A custom impact test bench underpinned the methodology used to quantify drone impact characteristics and develop safety governors. This bench utilized a linear catapult driven by an electric motor to propel drones into a stationary impact wall at controlled speeds. A propulsion trolley accelerated along a main rail, then released a drone-holding trolley onto a secondary rail immediately before impact, allowing for free rebound after collision.

The drone itself was mounted on a custom ultra-light carbon fiber support designed to minimize interference with the drone’s dynamic behaviour and overall inertia. Impacts were conducted at speeds ranging from 3 to 4m/s, representing a balance between typical cruising and scanning velocities for indoor micro-aerial vehicles.

This speed selection ensured measurable rebound distances within the 300mm rebound zone defined by the hardware. Force-time profiles were captured using three PCB Piezotronics load cells arranged in a triangular configuration behind the impact wall, providing comprehensive data on impact forces. Data-driven models were then generated to map pre-impact speed to both impulse and contact duration.

These models enable the direct computation of speed bounds necessary to maintain force limits specified by safety stakeholders. Scripts and a ROS2 node were developed to enforce these speed bounds online and simultaneously log compliance data, facilitating facility-specific policy implementation. The entire workflow was validated using multiple commercial off-the-shelf quadrotors, including the Bamboo Cognifly, Carbon Cognifly, DJI Avata, and Flywoo Flylens, and representative indoor assets. This validation demonstrated the preservation of task throughput while adhering to defined force constraints.

Impact force and duration characteristics of colliding drone configurations

Peak impact forces demonstrated a clear correlation with drone mass and structural compliance, with the bamboo Cognifly configuration registering the lowest force of 84.4 ±3.3 N. While this value falls below the ISO/TS 15066 neck impact threshold of 150 N, it still exceeds the facial impact limit of 65 N.

Conversely, the DJI Avata generated a peak force of 230.4 ±27.3 N, exceeding both the face and neck limits, though remaining below the 210 N back/shoulder threshold. The carbon Cognifly, utilising the same carbon fibre as the rigid Flywoo, achieved a 25% reduction in peak force, registering 105.6 N compared to the Flywoo’s 140.8 N, due to its TPU joints.

Contact duration remained largely independent of impact velocity across all tested configurations, with rigid designs like the Flywoo exhibiting stable durations of 17.2 ±0.2ms. More compliant structures, such as the bamboo Cognifly, showed higher variability in duration, measuring 41.4 ±11.3ms, attributable to nonlinear deformation behaviour.

Analysis revealed a trade-off between peak force and rebound energy, with configurations minimising peak forces tending to maximise rebound. The bamboo design, achieving the lowest peak force, exhibited the highest rebound energy of 16.4 ±2.9%. The research established a material-specific constant for contact duration, crucial for safety governor design, expressed as Fest = ∆p ∆tmaterial = m·v ∆tmaterial.

The study validated an end-to-end toolchain converting benchtop impact tests into deployable safety governors for drones, demonstrating preservation of task throughput while adhering to specified force constraints. The derived governors utilise a ROS 2 layer enforcing conservative speed limits when people are nearby, independent of upstream planning, and saturate /cmd vel to /cmd vel limited. This system fuses distance-aware ISO caps with impact-based caps, ensuring compliance with collaborative radii and guaranteeing empirically validated force limits for each airframe.

Impact Testing and Runtime Control for Safe Indoor Drone Operation

Researchers have developed a complete system for improving the safety of micro-aerial vehicles (MAVs) operating indoors near people. The work links physical impact testing with runtime safety controls, allowing for the creation of deployable safety governors for drones. A compact and reproducible impact rig was constructed to measure the force, impulse, and rebound experienced during collisions with drone airframes at typical indoor speeds.

Data from these impact tests were then used to create models that relate pre-impact speed to impulse and contact duration, enabling the computation of safe speed limits based on specified force thresholds. These limits are enforced by a software component, a ROS2 node, which logs compliance and supports facility-specific safety policies.

Validation across multiple drone platforms and in simulated environments demonstrated that the resulting governors maintain task performance while adhering to defined force constraints. The released datasets, scripts, and software form a practical toolchain for certifying indoor MAV operations. The current system has limitations including the restricted speed range of testing, which focused on rotor-off scenarios.

Future work will address rotor-on impacts, incorporate orientation compensation, integrate with model predictive control (MPC) systems, and conduct physical evaluations of the system’s performance. This measurement-backed approach provides a pathway toward safer and more certifiable indoor drone operations, offering a means to establish and enforce safety standards grounded in empirical data.