Roboat’s 3DFEA USV | UVD Klein Schiphorst adds: “As well as having many redundancies through this sensor arrangement, taking those sensor inputs and fusing them together lets us localise the ferry very accurately. Localisation via GNSS is a very wellestablished, tried and tested part of autonomous navigation, and we’ve stuck with the RTK-GNSS and IMU system from Swift Navigation as that continues to give us very good results. But, we will also be able to carry out some localisation using real-time Lidar modelling of landmarks along the Seine, just as we’ve done in Amsterdam’s canals in the past.” As well as reading the locations of objects relative to the boat, Lidar helps the ferry analyse and classify those objects so that it can determine the degree of avoidance manoeuvring necessary, thereby moving wider or holding still for longer in the face of a moving boat versus a moored one. The model of Lidar used on the ferry is an OS1 from Ouster, with Roboat specifically selecting a configuration with 1024 horizontal channels and 128 vertical channels for its resolution requirements. The Lidar measures with a range resolution of 0.1 cm and an angular sampling accuracy of ±0.01°. Roboat has close familiarity in working with such 360° spinning Lidars, and while it is also interested in trialling solid-state Lidars, obtaining these for testing has proved to be challenging. While the Lidars model objects around the ferry with close fidelity, each sensor module also integrates 6 MP cameras for defining and classifying those objects via the embedded computer vision. Klein Schiphorst says this may be more megapixels than the boat needs, but, given the fact that 35 souls may be carried per journey, excessive resolution, like excessive redundancy, was deemed the better side to err towards. Over the past four years, Roboat has advanced key aspects of its computer vision and related intelligent sensing tech, including improvements in the calibration of its sensors, the incorporation of select software and hardware innovations from autonomous road vehicles (though adapting them for marine use), and expansions of its body of AI training data through work with a growing number of public water-transport companies in Amsterdam and elsewhere. “We embed standard water maps into the ferry, so they know the limits of where the water is and isn’t, but most localisation is done in real time,” Klein Schiphorst notes. “And while there are rules and regulations on how the boat needs to move on water, they’re far more implicit than how cars need to move on roads, as boats don’t have to constantly deal with lane lines and traffic lights.” While using a DVL for USV navigation is unusual, it enables the ferry to detect current flows for aiding motion stabilisation, and also adds localisation information based on detections of the river’s bed and banks. The sensor hubs integrate OS1 Lidars from Ouster, as well as 6 MP cameras, which, along with other sensory inputs, ensure safety and redundancy in navigation and collision avoidance for the passengers onboard (Image courtesy of Roboat) Tired of generic job boards and speculative CVs? Why sift through irrelevant CVs when a laser focused recruitment portal can deliver quality over quantity? Contact [email protected] for more information on our 1x, 5x, 10x & unlimited job packages.
RkJQdWJsaXNoZXIy MjI2Mzk4