The historic pedestrian crossing is about to celebrate its 70th anniversary. It was George Charlesworth, later nicknamed “Doctor”. Zebra has come up with countless life-saving ideas over the last few decades. But modern urban lifestyles with new requirements call for alternatives to the time-honored white stripe. That’s where the new smart pedestrian crossing developed by George Carlsworth’s compatriot, the architecture firm Ambrelium. With its rather complex acronym, Starling (STigmaggie Ahadaptive R.Responsive L.obtaining intersection) It combines the latest AI and machine learning techniques to deliver a radically different experience than what we are used to. Installed in South London, the new prototype changes size, pattern and color to match the environment, including pedestrian and vehicle flow.
But what’s the main difference from traditional systems? The paint stripes are a mesh of LEDs connected to a neural data processing system fed by multiple cameras. In this way, smart crosswalks are always aware of pedestrians and vehicles around them and adapt to their needs. For example, when heading to work early in the morning, widen the crosswalk so that more people can cross safely. Alternatively, the placement of building doorways and subway stations may encourage pedestrians to cross diagonally. In such cases, the starling changes its pattern to match the natural flow of pedestrians. Basically, it’s the stigmage part of the acronym, which describes how individuals in large colonies of ants and other insects leave trails for other specimens to follow, providing new routes to food supplies and other points of interest. This is how you create a route. When there is no one around, the starling simply disappears.
However, organizing pedestrian flow is not the only special feature. If someone crosses the road while checking for the latest updates on their phone and gets lost from the crosswalk, LEDs highlight their location, giving drivers enough time to react . A similar thing happens when a child chases a ball across the road. A starling announces its position within 1/100th of her second. Finally, this smart intersection can also detect potential blind spots for bike riders and drivers and warn of pedestrians around them with specific light signals. This is achieved through machine learning that classifies objects in the scene. In this way, the system can predict the trajectory each will take and establish the safest point on the road to show the intersection.
The underlying technology includes two cameras, a neural processing unit, an ultra-bright LED panel and a steel frame supporting all cables, and a layer of resistant plastic that protects the entire system. The surface has been developed to improve grip and prevent slipping in wet weather, which will be welcomed by motorcyclists. The current prototype offers 22 square meters of reaction surface. The only thing left for pedestrians is to look away from their smartphone screens before crossing the road to reduce the number of road casualties.
ACO: Ant-based algorithms for AI
It might sound like an exaggeration to say that some of the most sophisticated AI algorithms in use today are based on scientific disciplines developed in the 1950s based on studying the behavioral patterns of ant colonies. . But stigma is like that. French biologist Pierre-Paul Grasset suggested it after scattering colonies of ants and termites. The point Grasse made was that there could be a decentralized system in which one agent could leave a specific trail of influence across the colony. This is the case for pathways that ants take to and from food sources. Each path is signaled by a pheromone, which evaporates unless reinforced by the passage of other specimens in the colony.
This is the origin of the so-called ant colony optimization algorithms used to solve problems like the traveling salesman. The algorithm must visit each city on its journey and find the shortest route back to the original city.
Source: Wired, Ambrelium
