Driverless

By combining disciplines Tianjin University has developed a variety of advanced algorithms in driverless perception, positioning, decision-making and control system.

Multi-layer channel

Multi-layer channel features combine deep neural network features with artificial ones. The average loss rate on the Caltech database is only 7.95 percent, and the detection speed is nearly 5 times higher than that of Fast-RCNN. These results are published in the top international journal IEEE Transactions on Image Processing.

         

Traffic sign database

Tianjin University established a Chinese traffic sign database with a total of 114 categories of 1 million traffic signs containing warnings, bans and other directions.Based on the improved Alexnet model, 99.31% accuracy of the German standard traffic sign dataset has been achieved; the accuracy of domestic traffic sign recognition is 98.37%.

World Intelligent Driving Challenge

Tianjin University participated in the 2017 World Intelligent Driving Challenge WIDC Driverless Team Competition held by the Tianjin Automobile Research Center and won the Excellence Award.

Unmanned roller compactor system

An on-site experiment of the unmanned roller compactor system was successfully carried out in the Lianghekou and Shuangjiangkou dams, realizing unmanned roller compacting and remote control.