Monitoring critical temperatures in permanent magnet synchronous motors is crucial for improving working reliability. Aiming at resolving the difficulty in online temperature estimation, an accurate and simple five-node lumped parameter thermal network (LPTN) is proposed and the mathematical model of the LPTN is built. Both radial and axial heat transfer paths inside the motor are considered to model the complete thermal circuit. In addition, an innovative parameter identification method based on multiple linear regression is applied to identify the parameters of the LPTN model. The parameters in the state equation are identified instead of the data of the motor, which are strongly dependent on the material and geometrical parameters. Finally, an open-loop estimation scheme based on the state equation and Kalman filter algorithm is adopted to predict the motor temperature online. The model performances are validated by extensive experiments under varying speed and torque conditions in terms of the accuracy and robustness. The results indicate that the temperature estimation error is within the range of ±5 °C in most cases and the proposed model can quickly follow the load variation. Besides, the online temperature estimation scheme and parameter identification method are easy and convenient to implement in an embedded system, which is feasible in automobile applications.

The total control range of asymmetric dual three-phase permanent magnet synchronous motor (ADT_PMSM) is divided into three different segments by voltage modulation range and current control dimensions and respective control strategies to get the injected voltages in harmonic subspace for three segments are presented in this paper. The three segments are sinusoidal current modulation region, sinusoidal voltage modulation region and overmodulation region. In sinusoidal current modulation region, resonant controller is adopted in the harmonic subspace to calculate the injected voltages, which can compensate the effect of six-phase voltage source inverter (VSI) dead time for two sets of three-phase windings and takes the pole correction into consideration. In sinusoidal voltage modulation region, open loop control strategy in harmonic subspace is adopted. Overmodulation region 1 and 2 are defined in overmodulation region, and the harmonic voltages which are injected into harmonic subspace to extend the modulation index are calculated based on superposition principle and VSD theory. In order to achieve smooth transition from different regions, a novel space vector pulse width modulation (SVPWM) technique for ADT_PMSM is proposed. The experimental results demonstrate the validity and feasibility of the suggested control approach.

The modeling of conduction and switching losses for insulated gate bipolar transistors (IGBTs) and free-wheeling diodes (FWDs) in automobile applications is becoming increasingly important, especially for the improvement of the system efficiency and the reliability prediction. The traditional modeling of conduction and switching losses based on the space vector pulse width modulation (SVPWM) is not applicable in practice due to the complex curve-fitting and the computation demands. In this paper, a simple and practical losses model for IGBTs and FWDs is proposed based on the SVPWM algorithm. Firstly, the traditional power losses model is introduced briefly. Then, the piecewise linear switching losses model and the conduction losses model based on the equivalent three-order harmonic model of the duty cycle are proposed. The comparison of experimental results between the traditional model and the proposed model is presented in the experiment validation. Furthermore, the power analyzer is adopted to measure the inverter losses, and the chips losses are further validated when other extra losses are considered. The proposed model shows good modeling accuracy with the large benefit of smaller measurement and lower computation requirements.

Active-short-circuit and locked-rotor modes are common abnormal operations in new energy vehicles. The IGBT junction temperature measurement for these two operating conditions is a challenging problem due to the unexpected large current and the asymmetric operation of semiconductor chips. In addition, different cooling flow rates have a significant influence on the heat dissipation, which will also have an impact on the building of the thermal model. Based on these difficulties, a modified Foster thermal network under active-short-circuit and locked-rotor modes has been presented considering different cooling conditions. The power loss models of the semiconductor chip under abnormal conditions are developed and a modified Foster thermal network based on the NTC temperature sensor is proposed. The model can be adapted to different cooling conditions since the thermal impedance fluctuates slightly at different cooling flow rates. The proposed thermal model is verified with inverter application under active-short-circuit and locked-rotor modes and the experimental performance shows good accuracy compared with the infrared camera measurement results.

The convolutional neural network (CNN) has been widely used in the field of self-driving cars. To satisfy the increasing demand, the deeper and wider neural network has become a general trend. However, this leads to the main problem that the deep neural network is computationally expensive and consumes a considerable amount of memory. To compress and accelerate the deep neural network, this paper proposes a filter pruning method based on feature maps clustering. The basic idea is that by clustering, one can know how many features the input images have and how many filters are enough to extract all features. This paper chooses Retinanet and WIDER FACE datasets to experiment with the proposed method. Experiments demonstrate that the hierarchical clustering algorithm is an effective method for filtering pruning, and the silhouette coefficient method can be used to determine the number of pruned filters. This work evaluates the performance change by increasing the pruning ratio. The main results are as follows: Firstly, it is effective to select pruned filters based on feature maps clustering, and its precision is higher than that of a random selection of pruned filters. Secondly, the silhouette coefficient method is a feasible method for finding the best clustering number. Thirdly, the detection speed of the pruned model improves greatly. Lastly, the method we propose can be used not only for Retinanet, but also for other CNN models. Its effect will be verified in future work.

Multi-Target tracking is a central aspect of modeling the environment of autonomous vehicles. A mono camera is a necessary component in the autonomous driving system. One of the biggest advantages of the mono camera is it can give out the type of vehicle and cameras are the only sensors able to interpret 2D information such as road signs or lane markings. Besides this, it has the advantage of estimating the lateral velocity of the moving object. The mono camera is now being used by companies all over the world to build autonomous vehicles. In the expressway scenario, the forward-looking camera can generate a raw picture to extract information from and finally achieve tracking multiple vehicles at the same time. A multi-object tracking system, which is composed of a convolution neural network module, depth estimation module, kinematic state estimation module, data association module, and track management module, is needed. This paper applies the YOLO detection algorithm combined with the depth estimation algorithm, Extend Kalman Filter, and Nearest Neighbor algorithm with a gating trick to build the tracking system. Finally, the tracking system is tested on the vehicle equipped with a forward mono camera, and the results show that the lateral and longitudinal position and velocity can satisfy the need for Adaptive Cruise Control (ACC), Navigation On Pilot (NOP), Auto Emergency Braking (AEB), and other applications.