Demo: Collision Prediction from Pairwise Ranging скачать в хорошем качестве

Demo: Collision Prediction from Pairwise Ranging

At the SPAN lab, we're developing new, more robust, collision prediction methods for autonomous vehicles and objects. Here Washington University in St. Louis student Jonathan Decavel-Bueff shows the performance of a pairwise collision prediction system that uses UWB measurements to measure the range over time between two objects, predicts future range, and predicts a collision if the two will collide soon. We use a platform able to collect range measurements at a very high rate because it performs frequency synchronization and timestamp data communication on a separate, narrowband radio. We implement our algorithm in firmware on a Nordic nRF Bluetooth microcontroller. We test it by putting the ranging platform on a robot chassis and showing that the algorithm detects collisions 0.5 sec before collision, and doesn't false alarm when the vehicles pass safely without collision. More information in the demo abstract: https://span.engineering.wustl.edu/pu...