A method and computer system for determining and controlling vehicle placement relative to roadway features such as lane markers, curbs or the roadway edge. The method comprises sending video 52 and light detecting and ranging (LIDAR) 50 data from sources affixed to a vehicle to a recurrent neural network (RNN) 54, which includes feedback between layers (see figure 3), and to a dynamic convolutional neural network (DCNN) 56 to identify a roadway feature. A softmax decision network 58 aggregates the outputs of the RNN and DCNN to determine the position of the vehicle on the roadway. This information is used to control the steering, braking and acceleration of the vehicle. The computer may extract a video frame image and a LIDAR frame image from the data (110, fig. 5; 210, fig. 6) and convert these into machine readable images (115, fig 5) for input into the RNN and DCNN. The method may also include comparing the vehicle position to a set of training data images to establish an error rate and determining and applying changes in weight for the RNN and DCNN from this error rate (see figure 5).
Vehicle lane placement
01.11.2017
Patent
Elektronische Ressource
Englisch
IPC: | G06V / B60W CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION , Gemeinsame Steuerung oder Regelung von Fahrzeug-Unteraggregaten verschiedenen Typs oder verschiedener Funktion / G05D SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES , Systeme zum Steuern oder Regeln nichtelektrischer veränderlicher Größen / G06T Bilddatenverarbeitung oder Bilddatenerzeugung allgemein , IMAGE DATA PROCESSING OR GENERATION, IN GENERAL / G01S RADIO DIRECTION-FINDING , Funkpeilung |
AUTOCONE: AN OMNIDIRECTIONAL ROBOT FOR LANE-LEVEL CONE PLACEMENT
British Library Conference Proceedings | 2020
|