Project description/goals

Based on real vehicle road driving status information, including engine speed, exhaust emissions, throttle opening, water temperature, vehicle speed, intake air volume, etc., this study judges whether the exhaust emissions exceed the standard and whether the vehicle load is overweight.

Importance/impact, challenges/pain points

• The data collection interval is 1 second, and the amount of data is huge, which increases the difficulty of data preprocessing.
• The road conditions are ever-changing, so that the state of the vehicle is complex and changeable. The model trained by this is likely to be under-fitting or unstable.

Solution description

We are going to classify exhaust emissions into two categories, either exceeding or not exceeding standards, and use machine learning methods to evaluate exhaust emissions. Secondly, this research uses a recurrent neural network (GRU) to build a predictive model, which makes the model memorable and can accurately predict the load of the car based on past historical information.

Key contribution/commercial implication

In this study, real-time monitoring of vehicle driving information is carried out to understand vehicle exhaust emissions and load information in a timely manner. Avoiding excessive emission of exhaust gas is of great significance to the protection of the ecological environment. Secondly, it reduces the overweight of vehicles and avoids road dangerous situations, which has an important impact on the safety of public transportation in our country.

Next steps

The previous work only trained one model. Without transferability, transfer learning or meta-learning methods will be used next, so that the trained model can be adapted to a variety of different car models.

Collaborators/partners

China Automotive Research Institute Automotive Inspection Center (Tianjin) Co., Ltd., School of Business Administration, Hunan University

Team/contributors

Lianmin Zhang，Chunling Wu ,Ke Zhou , Xiaoxin Bai ,Jinhui Cao ，Xinman Huang