Researchers at IIT Bombay, led by Professor Dwaipayan Mukherjee and scholar Chinmay Garanayak, have unveiled a groundbreaking control system that allows drone swarms to operate autonomously without GPS, inter-drone communication, or central command. This innovative method relies solely on onboard cameras, using ‘bearing-only’ measurements to track relative positions and maintain perfect formation.
This pioneering control scheme was specifically designed for Vertical Take-Off and Landing (VTOL) drones, known for their ability to launch without runways and hover, making them ideal for tasks in tight or challenging environments like surveillance. Professor Mukherjee emphasized the core of their innovation: ‘Autonomy in a swarm is a critical task. Our system empowers each drone to make decisions based purely on its own sensor data, removing the need for external information or human oversight. This approach fundamentally changes how drone swarms operate.’
With the ‘bearing-only’ control, each drone intelligently uses its camera to monitor nearby drones, calculating their precise bearing. Mr. Garanayak explained, ‘The entire formation control is achieved exclusively through these inter-agent bearing measurements.’ Crucially, this system functions completely independently, requiring no GPS, direct communication between drones, or any centralized computer control.
A significant advantage of this camera-centric approach is its resilience to noise, outperforming traditional distance sensors. This simplifies the drone’s hardware, reduces battery consumption, and lowers overall weight. Furthermore, the system excels in environments where GPS signals are absent or easily jammed, opening doors for discreet operations, including sensitive military assignments.
VTOL drones present a unique challenge as ‘underactuated’ systems, possessing six degrees of freedom but fewer direct control inputs. While they effortlessly manage vertical movement and rotation, their lateral and forward-backward motions require indirect manipulation. Professor Mukherjee noted, ‘Much of the existing research overlooks the complex underactuated dynamics of VTOL vehicles, focusing primarily on simpler kinematic models. This propelled us to delve into the full underactuated model of VTOL UAVs and investigate its potential for advanced formation control.’
Accurately controlling underactuated systems necessitates sophisticated dynamic models that account for position, orientation, velocities, forces, torques, and inertia. Prior efforts to implement bearing-only control on these complex models frequently faced issues with instability or outright failure under specific conditions. However, Mr. Mukherjee and Mr. Garanayak’s innovative control mechanism guarantees that drones will converge into and maintain their intended formation, even when initial positions are not perfect. Their work is backed by robust mathematical proofs, affirming the system’s unwavering reliability.
The research explores two primary operational scenarios. Initially, drones maintain a stable formation at a constant velocity, utilizing both bearing and bearing-rate data. For more dynamic situations where both formation and velocity fluctuate, the drones integrate their own velocity measurements alongside bearing data. This advanced system can manage diverse, time-varying configurations, enabling drones to traverse confined spaces, swiftly switch to single-line formations, and seamlessly adapt to evolving mission demands.
The next crucial step for the researchers involves experimental testing of their control scheme with a live drone swarm. Looking ahead, a key objective is to develop collision avoidance systems backed by theoretical guarantees. ‘Many current algorithms use arbitrary collision avoidance methods without proven reliability,’ stated Mr. Mukherjee. ‘Our goal is to resolve collision avoidance, both with environmental obstacles and among drones, through a robust, theoretically sound approach.’