Mobility models are pivotal for the analysis and effective simulation and modelling of drone swarms, and mobile ad hoc networks in general. In the paper entitled “Swarm Mobility Models and Impact of Link State Awareness in Ad Hoc Routing”, supported by the RAINBOW project, the authors analyze the impact of link-state awareness in ad hoc deployments for various swarm mobility models. There exist several such models to approximate the movement of nodes in a swarm network.
- Gauss-Markov mobility model: this model attempts to accommodate various mobility randomness levels. This model assumes that velocity vectors of various swarm nodes are completely independent. Swarm nodes are initially assigned a defined direction and velocity. At constant intervals, both velocity and direction are updated using a randomizer, also considering previous velocity and direction. Consideration of previous node velocity and direction is typically successful in mitigating possible abrupt path alterations. This model is mature and well-researched in the context of drone swarms; there also exists a variant of this model called 3D Gauss-Markov, specifically developed for multi-altitude FANETs, as it considers and randomizes mobility in all three dimensions.
- Semi Random Circular mobility model: this model attempts to accommodate curved and circular trajectories. This makes the model applicable in the context of patrolling and inspection applications, and military/repetitive surveillance operations in general security. The semi-random circular mobility model has proven to be more efficient than existing models for the simulation of curved maneuvering, since it is the first one specifically designed for curved scenarios.
- Random Waypoint mobility model: this model is capable of simulating node motion based on linear motion and its derivatives (turns, stops). It considers and models the location, velocity and acceleration change of a swarm node. In this model, each node defines a random destination, within a predefined grid, and then engages it at a random velocity. Should the node achieve this defined goal, it pauses for a random amount of time (within predefined constrains), and then defines a new destination. The process then repeats itself. Thanks to its low computational complexity and overall algorithmic simplicity, it is one of the most popular mobility models for mobile ad hoc networks.
- Particle Swarm mobility model: this model tries to maintain a collision-free swarm node distribution at all times by considering the spatial relationship amongst networked nodes. The first step implemented by this algorithm is the logging of the node velocities and waypoints. Then, for each node, the model generates new velocity vectors and waypoints. Lastly, the particle swarm mobility model performs various adjustments in the velocity vector of swarms nodes, to avoid collisions, and the process is repeated for every waypoint. This model succeeds in keeping all UAV nodes in safe distances, while achieving high temporal and spatial correlation and guaranteeing path availability \cite{XianfengLi2017}. This makes the model suitable for the modelling of intelligent swarms, composed of predicatively maneuvering nodes.
- Paparazzi mobility model: this model incorporates a total of five possible node maneuvers, namely “stay-at”, “waypoint”, “eight”, “scan” and “oval”. Combinations of those five aforementioned basic maneuvers are capable of covering virtually all realistic node movements in the three-dimensional space. This model is ideal for simulating maneuvering of multi-node swarms. It provides an accurate description of a swarm’s mobility in a real-life environment thanks to the combination of a realistic number of commonly-used maneuvers.
Each mobility model has its own advantages, disadvantages and unique identifying characteristics. In the paper presented at CSNDSP 2022 in Porto, by George Amponis (K3Y Ltd), conclusions were drawn in terms of peak performance (measured as quality of service, and overall throughput), depending on the utilized routing protocol, and the considered mobility model, depending on the target use case.
