This system helps a wheeled robot navigate a physical space more accurately. It uses two cameras and an inertial sensor to identify objects around it. By predicting where objects should be based on its own motion and comparing that to what the cameras actually see, the robot can distinguish stationary objects from moving ones, using only the stationary ones to build its map and figure out its own position.
Why it matters: Since 2019, advancements in deep learning and edge AI have made real-time, robust identification and segmentation of moving objects far more practical. This could significantly enhance the accuracy and efficiency of filtering moving features for SLAM, which this invention aims to do.
AI gives you a few directions you could take this. Pick one, and we check whether your version is different enough to patent, then write the filing.
Reinvent this with AI