Don't Let the Past Slow You Down
WiSwarm middleware prioritizes data freshness to keep time-sensitive applications running at top speed over standard Wi-Fi networks.
As computational resources simultaneously increase in power and decrease in size, more and more use cases are emerging for distributed networks of devices. Whether it is a swarm of drones on a search and rescue mission, preventative maintenance monitors on industrial equipment, or smart city intersections that are in communication with self-driving cars, many of these applications are highly time sensitive. And while modern advances in processing units and memory are, by and large, keeping up with needs, traditional wireless networks may not be up to the job.
Computer vision applications that transmit images to a central server for processing, in particular, can place a strain on traditional Wi-Fi networks. All the more so when those drones are flying fast and need quick responses from the controlling system to keep them on track — and keep them from crashing. This can be dealt with by using special radio equipment, but that adds bulk to systems, which is undesirable in many cases, such as with drones, where keeping weight down is crucial. But with Wi-Fi networks being so pervasive, it would be much more desirable to leverage them. If they could handle the load, that is.
Unfortunately, these networks were not designed with time sensitive applications in mind. As it turns out, Wi-Fi networks may actually be up to the task — with a little help, anyway. A team composed of researchers from MIT and Columbia University have recently developed a system called WiSwarm that can prioritize the transmission of relevant information. This allows for the communication of much fresher data than would be possible under normal circumstances.
Under most circumstances, dropping packets is not a good thing. If your Wi-Fi router were to discard older packets while loading this webpage, well, it would not make for a great experience. But in some cases, all that matters is the latest information. When a drone transmits images for navigation, the only thing that matters is where it most recently was. Where the drone was at any point in time prior to that is no longer relevant, so transmitting that data would only serve to waste network resources.
WiSwarm was designed with these types of applications in mind. It is a piece of middleware that sits between connected devices and a traditional Wi-Fi router. Since the middleware was implemented at the application layer, it can interface with a standard 802.11 Wi-Fi router. As packets come in, they are inspected and stale packets that are no longer useful are discarded, the most relevant information is prioritized, and collisions between packets are prevented.
An age-of-information metric is calculated that determines the freshness of all incoming data from the perspective of the destination. This metric helps determine which connected devices are working off of the most stale data, and in turn prioritizes them for receiving updates. This is factored in with a priority score given to each device — a fast moving drone would need more frequent updates than a slower moving drone, so it would have a higher priority score. The drone’s channel reliability and the amount of time since the drone sent in an update are also included in the final determination as to which device’s communications are to be transmitted at any given time.
The WiSwarm system was tested with the help of multiple drones that were tasked with tracking objects in motion. An onboard camera transmitted images to the middleware, which would then decide when to forward them on to a central server that was providing each drone with the navigation instructions to carry out its task. When two drones were in the mix, WiSwarm kept the incoming data two times fresher, which made the tracking capability six times better. When the number of drones was increased to five, Wi-Fi alone could not keep up and the drones lost their way. Using WiSwarm, however, allowed all five drones to keep on track without any problems.
With a bit more refinement, the researchers believe that WiSwarm will be helping to power swarms of drones in search and rescue operations, in building inspections, and much more in the near future.