Swarm Intelligence

In our last blog, we read about the origins, applications and configurations of swarm drones. Today we bring a new article explaining swarm intelligence and probable ways to counter swarm of drones.

Drone swarms encompass the integration of networking and computational systems, which together contribute to the closed-loop operation of the Networked Control System. Sensing acts to gather data from the environment into the swarm system while the networking system ensures that such data gets to the points on the network (swarm) where computation should be done, as well as ensuring that computational results are delivered to the drones that require that information to take actions needed to adapt the overall behavior of the swarm.

Communication System

The networking system builds a communication graph for data to be exchanged between drones, and between drones and ground infrastructure. The three elements i.e., sensing, decision-making and actuation may reside in different drones, providing a cooperative sensing.

For instance, a drone may sense a new obstacle that does not interfere with its own position, but it may have an impact on the overall swarm. This information is made available to another drone able to compute an action, which may be executed by a set of other drones that are in the way of such an obstacle.

Fading and interference mitigation is addressed by wireless beam forming, and use of multiple dual-polarized and cross-polarized MIMO configured light-weight antenna systems. Multi-hop drone design implementation again helps in avoiding fading while contributing to spectrum efficiency and increased transmission coverage. Latency requirements are expected to be met in commercial civilian applications through 4G networks presently and subsequently 5G. A typical packet size for UAV communications is between 17 and 263 bytes. While 4G speeds are sufficient for these packets, 5G will allow for additional data streaming including data types, such as video from payload cameras or data from payload light detection and ranging (LiDAR) systems. However, packet loss and the performance of orthogonal frequency-division multiplexing for UAV communication have been analyzed, and with increased speeds and infrastructure updates of 5G systems, performance will increase. For military drone swarms, Direct Sequence/ Frequency Hopping Multiple Access schemes will ensure Low Probability of Intercept of the flying ad hoc wireless network.?

Computation System

The computation system involves decision making. In the non-interactive deployment, the decision making can be done by the drones themselves in a distributed manner, and by a centralized control entity.?However, in a swarm of drones with high autonomy level, the decision-making can be done by the drones themselves. Centralized decision-making process offers simple solutions in terms of the overall system design, while contributing to reduce the energy consumption of each drone. On the other hand, a distributed decision-making approach may lead to a more robust and scalable drone swarm.

The true strength of a swarm lies in its ability for emergent behaviour also called swarm intelligence. The swarm capability is not just the mere sum of its individual drone capabilities. In this case algorithms mimic biological systems such as flocks of birds, ants and honeybees where individual agents follow simple rules. Particle Swarm Optimization (PSO) algorithm mimics birds and is used for path planning, and moving target searching by continuously updating velocities and positions of the drones in the solution space. By sharing this information, the algorithm allows different drones to move towards best positions relative to their neighbours to maintain strong bonds for collective sailing in attaining optimum performance.

Bee Colony Optimisation (BCO) mimics the behaviour of bees that have different roles in a colony. Hence, this algorithm may have a good applicability to the optimization of swarms encompassing drones with different roles. Ant Colony Optimisation (ACO) mimics the behaviour of ants which go in search for food. They drop pheromones as a guide to ants that are following them. A larger number of pheromones indicate a likely food source. The algorithm has strong cooperative behaviors and so have strong global optimal abilities as well as being flexible to implement.

Challenges

China’s ability to design a single chip housing the flight control system, mission planning, intelligent decision-making, and dynamic networking between drones, as well as the ability to recognize targets and other objects manufactured by China Electronic Technology Group Corporation was responsible for its successful tactical implementation of the drones swarm. This is similar to the special-purpose swarm-enabled version of Intel’s?Movidius AI processor, which has also been used to control smart drones. This development is likely to revolutionise swarm development since the costs will substantially reduce and may well find its way to terrorist groups.

Countering a swarm is quite difficult due to their inherent survivability and resilience. RF detection and brute force jamming is an option. However, since role based drones are likely to be invested with considerable autonomy it would be difficult to neutralise them entirely. Laser weapons since they are not area weapons but pinpoint directional weapons are also likely to be ineffective. Micro missiles which fire multiple munitions over an area would be effective. Artillery fire of air burst ammunition could be effective. The best method is likely to be to launch a counter swarm tasked with destruction of the hostile swarm.