The majority of the more recent studies have mainly focused on how to achieve energy-efficient goals by optimizing the driving behavior for human-driven vehicles or designing trajectory planning and tacking algorithms for autonomous vehicles. However, the energy-saving advantages of autonomous vehicles have not been quantitatively and theoretically explored. Therefore, this study aims to specifically clarify whether autonomous vehicles use less fuel than human-driven ones. First, the differences in driving behavior, regarding speed control, between autonomous vehicles and human-driven vehicles were compared. The most notable difference between them is that an autonomous vehicle can control the vehicle speed more effectively, with less speed fluctuations than a human driver. Subsequently, the influence of speed fluctuation on vehicle fuel consumption (L/100 km) was formulated based on the vehicle specific power (VSP) model. The mathematical deduction showed that the fuel consumption is proportional to the speed fluctuation under the same mean speed. Finally, simulation experiments were conducted under real scenarios. The simulation data showed that the fuel consumption increases almost linearly with the increase in speed fluctuation. Field experiments were also conducted on the fuel consumption of an autonomous vehicle under different driving modes. The experimental data showed that the fuel consumption also increases almost linearly with the increase in speed fluctuation. In the human-driven mode, the fuel consumption increased by 5.6% and 14.7%, respectively, compared with that in the autonomous mode at average speeds of 20 km/h and 40 km/h. Furthermore, the maximum fuel consumption was up to 60% more when the autonomous vehicle was driven by a driver, as the driving behavior displayed greater speed fluctuations.

The discriminative correlation filter (DCF) is commonly utilized in UAV tracking because of its high tracking accuracy and computing speed. However, in aerial tracking scenarios, challenges such as target occlusion and similar object interference are likely to cause the predicted object position to deviate from the correct motion trajectory. To alleviate this issue, this paper proposes a correlation filter algorithm based on trajectory correction and context interference suppression for real-time aerial tracking. First, a tracking quality evaluation metric is proposed to determine the confidence of the current tracking results. When the object is in a low confidence status, the state matrices of the object position and velocity are constructed, and the Kalman filter strategy is utilized to correct the tracking trajectory automatically. In addition, temporal context-response regularization is designed to fully exploit previous temporal information in order to suppress background interference. Extensive experimental results on four mainstream datasets demonstrate that the proposed algorithm has high tracking performance while achieving a real-time tracking speed of 32 fps on a single CPU.

Conventional fuel consumption prediction (FCP) models using neural networks usually adopt driving parameters, such as speed and acceleration, as the training input, leading to a low prediction accuracy and a poor correlation between fuel consumption and driving behavior. To address this issue, the present study introduced jerk (an acceleration derivative) as an important variable in the training input of four selected neural networks: long short-term memory (LSTM), recurrent neural network (RNN), nonlinear auto-regressive model with exogenous inputs (NARX), and generalized regression neural network (GRNN). Furthermore, the root-mean-square error (RMSE), relative error (RE), and coefficient of determination (R²) were used to evaluate the prediction performance of each model. The results from the comparison experiment show that the LSTM model outperforms the other three models. Specifically, the four selected neural network models exhibited an improved accuracy in fuel consumption prediction after the jerk was added as a new variable to the training input. LSTM exhibited the greatest improvement under the high-speed expressway scenario, in which the RMSE decreased by 14.3%, the RE decreased by 28.3%, and the R² increased by 9.7%.