Source-linked AI summary
Outdoor flocking and formation flight with autonomous aerial robots
Gábor Vásárhelyi, Csaba Virágh, Gergő Somorjai, Norbert Tarcai, Tamás Szörényi, Tamás Nepusz, Tamás Vicsek
TL;DR
The paper addresses the limited experimental evidence for decentralized autonomous aerial flocks operating outdoors under noisy, delayed, and error-prone conditions. It develops a locally communicating, onboard-controlled flocking framework and demonstrates stable outdoor flight with up to 10 robots across flocking and formation tasks. The authors also use realistic simulations to examine delay robustness and scalability to larger flocks.
Problem
Experimental evidence for autonomous aerial formation flight is limited, especially for decentralized outdoor systems operating with noise, delay, and communication errors.
Method
The system combines decentralized onboard flocking control with local broadcast communication, bio-inspired modeling, and realistic simulations for optimizing behavior under delays.
Results
The system achieved stable decentralized autonomous outdoor flight with up to 10 robots, including flocking, object avoidance, target tracking, and rotating-ring, grid, and line formations.
Takeaways & Limitations
Local neighbor-based communication and decentralized onboard control supported stable collective flight and were demonstrated in simulations to scale to much larger flocks.
Takeaways & Limitations
The current algorithms are two and a half dimensional: flocking and formation flights occur in two dimensions, while altitude changes are limited to individual take-off and landing.
Abstract
from arXiv · showhide
We present the first decentralized multi-copter flock that performs stable autonomous outdoor flight with up to 10 flying agents. By decentralized and autonomous we mean that all members navigate themselves based on the dynamic information received from other robots in the vicinity. We do not use central data processing or control; instead, all the necessary computations are carried out by miniature on-board computers. The only global information the system exploits is from GPS receivers, while the units use wireless modules to share this positional information with other flock members locally. Collective behavior is based on a decentralized control framework with bio-inspiration from statistical physical modelling of animal swarms. In addition, the model is optimized for stable group flight even in a noisy, windy, delayed and error-prone environment. Using this framework we successfully implemented several fundamental collective flight tasks with up to 10 units: i) we achieved self-propelled flocking in a bounded area with self-organized object avoidance capabilities and ii) performed collective target tracking with stable formation flights (grid, rotating ring, straight line). With realistic numerical simulations we demonstrated that the local broadcast-type communication and the decentralized autonomous control method allows for the scalability of the model for much larger flocks.
A. Basic Unit
The flying unit uses a partially open-source quadcopter with an extension-board control interface, supporting automatic operation while retaining manual interruption. Each unit weighs around 1 kg, spans 80 cm tip-to-tip, and flies for up to 20 minutes.
- The base MK Basicset L4-ME quadcopter provides onboard self-stabilization and altitude hold, while originally supporting manual R/C control.
- The main board accepts virtual pitch, roll, yaw, and altitude commands from an extension board for full automatic control.Manual interruption through the remote controller remains available.
- Around 1 kg take-off mass and 80 cm tip-to-tip diameter characterize each unit.
- 20 minutes is the maximum flying time achieved using 3000mAh, four-cell LiPo batteries.
B. Processor Board
The custom FlockControl extension board integrates sensing, GPS, wireless communication, and onboard computation, connecting to the copter’s main board through SPI.
- The FlockControl board combines a 3D gyroscope, accelerometer, magnetometer, pressure sensor, GPS receiver, 2.4GHz XBee unit, and Linux mini-computer.
- The extension board receives attitude, heading, battery, remote-control, and error-state information from the base unit through its connection to the main board.
- Flock members exchange ID, position, velocity, attitude, and status information through the board’s XBee communication link.
C. Communication
Communication is local and broadcast-based: each robot sends packets without one-to-one links or a mesh network, then processes messages from nearby flock members. This design supports scalability while operating under substantial communication delays and occasional outages.
- C. Communication: Broadcast transmission lets units send data without establishing one-to-one connections or a mesh network.
- C. Communication: Robots process packets only from neighbors within a typical 50–100 m communication range, mirroring local interaction in animal flocks.
- C. Communication: Local communication is presented as a simple means of keeping the network scalable to larger flock sizes.
- C. Communication: 0.4s ± 0.2s is the typical communication delay with ten active units, while random outages lasting seconds can also occur.The delay spans GPS reception, XBee broadcast, and XBee reception on another unit.
D. Velocity Control
The system combines decentralized flocking with onboard velocity control and simulation-based optimization to maintain stable autonomous outdoor flight despite delays, noise, and disturbances. Experiments demonstrated stable flocking, target tracking, multiple formations, and self-organized motion with up to 10 robots.
- Velocity control: Two independent horizontal PID controllers convert flocking velocity outputs into body-fixed steering signals, while altitude hold maintains a predefined height.North and East control axes are treated independently and transformed using heading information from onboard sensors.
- Velocity control: Feed-forward bias reduces the integral controller’s burden and improves response when tight flocking requires rapid adjustments.The bias predicts the required adjustment from prior experience as measured velocity approaches the target.
- Velocity control: 1.5 s overall time lag remained during stable tracking in moderate wind up to 5 m/s, with delays arising mainly beyond the PID loop.Base-unit reaction time, low GPS update rate, and copter inertia contributed substantially to the lag.
- Velocity control: Delay-sensitive flock dynamics can generate chaotic oscillations, so the flocking algorithm was optimized for stable, overshoot-free but slower control.The authors compare these delay-driven oscillations with ghost traffic jams caused by slow human reaction times.
- Experimental demonstrations: Formation stability was supported by closest-neighbor-distance standard deviations as small as 1 m over maintained 1–10 minute periods.The system also allowed real-time changes to equilibrium distance and formation shape through the R/C controller.