Novel scheme by IIT Bombay researchers to control drones can enable complex formation flying using only camera data, without GPS or inter-drone communication.

Vertical take-off and Landing (VTOL) unmanned aerial vehicles (UAVs), such as drones used for photography, can just lift off the ground and fly, without needing a runway. They can hover and stop mid-air. They are thus very suitable for monitoring, surveillance and other tasks that require confined-space movement. 

Prof Dwaipayan Mukherjee and research scholar Chinmay Garanayak from the Indian Institute of Technology Bombay have proposed a control scheme that enables VTOL unmanned aerial vehicles in swarms to maintain intended formations and navigate by merely ‘looking’ at their immediate neighbours. They have also provided a mathematical proof for the stability of their proposed control scheme. 

Most currently used formation and navigation control systems rely on an external positioning system, such as GPS, or on a human operator or a central computer to control the movements of member drones. “Autonomy in a swarm is a critical task. This means that vehicles in a swarm should be able to decide their ‘actions’ based on variables they can measure with their on-board sensors, instead of having to rely on some global information being fed to them or some human/centralised computer deciding what their action ought to be. This is where our paradigm differs from usual ones”, says Prof Mukherjee. 

Garanayak and Mukherjee have proposed a ‘bearing-only’ control scheme in which each member drone uses its relative position with respect to its immediate neighbour drones (bearing information) to move in the intended direction while maintaining its position in the swarm. “In bearing-only control, the goal is to achieve formation control using only interagent bearing measurements,” say the researchers. 

A member-drone can use its on-board camera to capture and calculate the bearing information. It does not need to use GPS or communication from a central computer or neighbouring drones. Camera-based measurements are generally less prone to noise than conventional distance sensors, simplifying the drone’s sensor system, reducing the drone’s battery requirements and overall weight. The scheme can work well in areas where GPS is not available or communication could be jammed, thus making the system efficient, reliable, robust, and autonomous. It makes stealth-mode operations easier, making it useful for covert military operations. 

VTOL vehicles are controlled by multiple rotors, usually on top of the vehicle. A drone can move up-down, left-right and forward-backwards, tilt upward-downward, left-right, and turn left-right. Since it can move in these six ways, it is said to have six degrees of freedom. However, VTOL drones have mechanisms to directly control only their up-down movement and rotational motions around the three axes. The left-right and forward-backwards movements need to be indirectly controlled by carefully calculating available direct control mechanisms. Such systems, in which the number of directly controlled movements is fewer than the degrees of freedom, are called underactuated systems. It is not easy to control underactuated systems, and the required control system is complex. “Many of the results in the literature do not address the underactuated dynamics of VTOL vehicles and only focus on the kinematic model. This motivated us to consider the fully underactuated model of the VTOL UAV and explore its applicability to formation control,” explains Prof Mukherjee. 

Underactuated systems need to be modelled using dynamic models that include position, orientation, velocities, as well as forces, torques and inertia information. Previous attempts to apply bearing-only control to these dynamic models often fell short. The stability proofs failed for some of the earlier solutions, while the control mechanisms for some broke down in certain situations. Garanayak and Mukherjee have meticulously developed a new control scheme that guarantees stability. Under its control, the drones will reliably converge to and maintain their desired formation, even when starting from imperfect positions. They have also provided rigorous mathematical proofs that their control scheme achieves stability.

Garanayak and Mukherjee’s work addresses two main scenarios: one in which the drones need to maintain formation at constant velocity, and the other in which the formation and velocity change over time. In constant-velocity scenarios, drones use only the bearings and the rate at which they change (bearing rates) to maintain their formation. For the more complex time-varying scenarios, where the formation might need to change shape or the leader drone might be accelerating or turning, the drones also incorporate their own velocity measurements in addition to bearing measurements. A key improvement of Garanayak and Mukherjee’s scheme over prior research is the ability to handle arbitrary time-varying configurations. This is vital for real-world applications where drones may need to navigate narrow passages, temporarily reconfigure into a single line, or adapt to changing mission requirements.

Garanayak and Mukherjee provide a solution, strongly backed by theory, to a practical problem. They plan to test their control scheme experimentally using a drone swarm. On their future roadmap, Prof Mukherjee informs, “Most existing algorithms rely on ad-hoc collision avoidance schemes that do not come with any theoretical guarantees. Collision avoidance with objects in the environment and among drones is a challenge we are trying to tackle at a theoretical level.”

Funding Information : This work was supported in part by ANRF funded project with project code: CRG/2023/002280.

In Hindi In Marathi