QCNet: Revolutionizing Autonomous Vehicle Safety with Advanced Trajectory Prediction

There has been a large-scale transformation in Automobiles from normal vehicles to electric vehicles. Now, This is being transformed into more advanced automobiles called self-driven automobiles. This is done with the help of Advanced Learning Algorithms of Artificial Intelligence and Machine Learning. Researchers at the City University of Hong Kong have made a new AI system for self-driven automobiles. This model can also predict whether the pedestrians are nearby or not and also predicts where the nearby automobiles and pedestrians can go accurately. It also makes this much faster, making it safe for self-driven automobiles. Researchers also said that getting these predictions correct is very important as even a small change in prediction can cause a worse accident. The problem with the existing solutions is that these lack in giving accurate predictions.

To solve this problem, a team of researchers developed a groundbreaking AI system called QCNet. This improves the prediction of vehicle and pedestrian movements in self-driven automobiles. This model works in real-time and also provides us with the limitations of the existing models. This relies on the concept of relative space-time positioning. These properties make it capable of understanding the traffic rules and interactions with other people. This also enables it to predict future trajectories that comply with maps and avoid collisions. For the evaluation of the model, the researchers used large datasets like Agroverse1 and Agroverse2. These datasets contain a vast amount of autonomous driving data and high-definition maps from various U.S cities. These datasets are also called challenging benchmarks for behavior prediction.

The research tested the model and found that both the speed and accuracy were quite good. The testing was an average of Agroverse 1 and Agroverse 2. The model also took more than six seconds for some of the predictions, but the predictions were accurate. In complex traffic analysis, which involves a large number of road users and map polygons, the accuracy of the model was about 85%.

Researchers are still working on applying this model to predict human behavior, which will also decide whether the model is efficient or not. This whole process falls under the category of Image Processing as well as Computer Vision. Researchers said that the model still has some problems related to predictions and self-driving efficiency, which will be improved further by hyperparameter testing. This was one of the most significant research in the history of Automobiles.


Check out the Paper and Reference ArticleAll Credit For This Research Goes To the Researchers on This Project. Also, don’t forget to join our 30k+ ML SubReddit, 40k+ Facebook Community, Discord Channel, and Email Newsletter, where we share the latest AI research news, cool AI projects, and more.

If you like our work, you will love our newsletter..

Bhoumik Mhatre is a Third year UG student at IIT Kharagpur pursuing B.tech + M.Tech program in Mining Engineering and minor in economics. He is a Data Enthusiast. He is currently possessing a research internship at National University of Singapore. He is also a partner at Digiaxx Company. 'I am fascinated about the recent developments in the field of Data Science and would like to research about them.'

🐝 Join the Fastest Growing AI Research Newsletter Read by Researchers from Google + NVIDIA + Meta + Stanford + MIT + Microsoft and many others...