Optimizing Steering Angle Predictive Convolutional Neural Network for Autonomous Car

Deep learning techniques,particularly convolutional neural networks(CNNs),have exhibited remarkable performance in solving visionrelated problems,especially in unpredictable,dynamic,and challenging environments.In autonomous vehicles,imitation-learning-based steering angle prediction is viable due to the visual imagery comprehension of CNNs.In this regard,globally,researchers are currently focusing on the architectural design and optimization of the hyperparameters of CNNs to achieve the best results.Literature has proven the superiority of metaheuristic algorithms over the manual-tuning of CNNs.However,to the best of our knowledge,these techniques are yet to be applied to address the problem of imitationlearning-based steering angle prediction.Thus,in this study,we examine the application of the bat algorithm and particle swarm optimization algorithm for the optimization of the CNN model and its hyperparameters,which are employed to solve the steering angle prediction problem.To validate the performance of each hyperparameters’set and architectural parameters’set,we utilized the Udacity steering angle dataset and obtained the best results at the following hyperparameter set:optimizer,Adagrad;learning rate,0.0052;and nonlinear activation function,exponential linear unit.As per our findings,we determined that the deep learning models show better results but require more training epochs and time as compared to shallower ones.Results show the superiority of our approach in optimizing CNNs through metaheuristic algorithms as compared with the manual-tuning approach.Infield testing was also performed using the model trained with the optimal architecture,which we developed using our approach.

Human steering angle estimation in video based on key point detection and Kalman filter

Human pose recognition and estimation in video is pervasive.However,the process noise and local occlusion bring great challenge to pose recognition.In this paper,we introduce the Kalman filter into pose recognition to reduce noise and solve local occlusion problem.The core of pose recognition in video is the fast detection of key points and the calculation of human steering angles.Thus,we first build a human key point detection model.Frame skipping is performed based on the Hamming distance of the hash value of every two adjacent frames in video.Noise reduction is performed on key point coordinates with the Kalman filter.To calculate the human steering angle,current state information of key points is predicted using the optimal estimation of key points at the previous time.Then human steering angle can be calculated based on current and previous state information.The improved SENet,NLNet and GCNet modules are integrated into key point detection model for improving accuracy.Tests are also given to illustrate the effectiveness of the proposed algorithm.

Research on Steering Resistance Moment of Wheel Loader

The wheel loader as the research object in present article,its steering mechanism is analyzed for the relationship between the steering cylinder displacement and the steering angle,which means,the relationship between the arm of steering resistance moment and the steering angle.In addition,the relationship between the in-situ steering resistance moment and the wheel angle is also be analyzed by integrating the interaction between the tire and the ground.The Matlab will help to build the mathematical modeling for verification.

Ultrasonic beam steering using Neumann boundary condition in multiplysics

The traditional one-dimensional ultrasonic beam steering has time delay and is thus a complicated problem. A numerical model of ultrasonic beam steering using Neumann boundary condition in multiplysics is presented in the present paper. This model is based on the discrete wave number method that has been proved theoretically to satisfy the continuous conditions. The propagating angle of novel model is a function of the distance instead of the time domain. The propagating wave fronts at desired angles are simulated with the single line sources for plane wave. The result indicates that any beam angle can be steered by discrete line elements resources without any time delay.

ACTIVE FRONT STEERING DURING BRAKING PROCESS

An active front steering （AFS） intervention control during braking for vehicle stability is presented. Based on the investigation of AFS mechanism, a simplified model of steering system is established and integrated with vehicle model. Then the AFS control on vehicle handling dynamics during braking is designed. Due to the difficulties associated with the sideslip angle measurement of vehicle, a state observer is designed to provide real time estimation. Thereafter, the controller with the feedback of both sideslip and yaw angle is implemented. To evaluate the system control, the proposed AFS controlled vehicle has been tested in the Hardware-in-the-loop-simulation （HILS） system and compared with that of conventional vehicle. Results show that AFS can improve vehicle lateral stability effectively without reducing the braking performance.

Theoretical Study of Double Cost Function Linear Quadratic Regulator(LQR)

Double cost function linear quadratic regulator (DLQR) is developed from LQR theory to solve an optimal control problem with a general nonlinear cost function. In addition to the traditional LQ cost function, another free form cost function was introduced to express the physical need plainly and optimize weights of LQ cost function using the search algorithms. As an instance, DLQR was applied in determining the control input in the front steering angle compensation control (FSAC) model for heavy duty vehicles. The brief simulations show that DLQR is powerful enough to specify the engineering requirements correctly and balance many factors effectively. The concept and applicable field of LQR are expanded by DLQR to optimize the system with a free form cost function.

Digital Beamforming by Multiple Chirp Zeta Transforms for 3-D Sonar Imaging Systems

The computational load is prohibitive for real-time image generation in 3-D sonar systems, particularly when the steering angle approximation is required. In this paper, a novel multiple Chirp Zeta Transforms （MCZT） beamforming method in frequency domain is being proposed. The single long-length Chirp Zeta Transform （CZT） in the original CZT beamforming is replaced by several CZTs with smaller lengths for different partitions along each dimension. The implementing routine of the algorithm is also optimized. Furthermore, an avenue to evaluate the estimating error for the angle approximation in 3-D imaging applications is presented, and an approach to attain valid partitions for the steering angles is also flhistrated. This paper demonstrates a few advantages of the proposed frequency-domain beamforming method over existing methods in terms of the computatianal complexity.

Method for measuring the steering wheel angle of paddy field agricultural machinery by integrating RTK-GNSS and dual-MEMS gyroscope

Aiming at the application environment of paddy agricultural machinery with bumpy and undulating changes,the problems affecting the method for steering wheel angle measurement by MEMS gyroscope were analyzed,and a wheel angle measurement method combining Dual-MEMS gyroscope(dual MEMS gyroscope)and RTK-GNSS was designed.The adaptive weighting method was used to fuse the heading angle differentiation of RTK-GNSS,the MEMS gyroscope angle rate,and velocity data,and the rod-arm compensation was performed to accurately obtain the angle rates of the body and steering wheels of agricultural machinery;the difference between the combined angular rate of the steering wheel of the agricultural machinery and the angular rate of the agricultural machinery body was obtained,and the integrator is used to integrate the difference to get the wheel steering angle value,and the Kalman filter was designed to make feedback correction for the integration process of angle calculation to eliminate the errors caused by the gyroscope zero bias,random drift,and gyroscope rod arm effect,and to obtain the accurate value of wheel steering angle.A comparative test with the connecting rod wheel angle sensor was designed,and the results show that the maximum deviation is 4.99°,the average absolute average value is 1.61°,and the average standard deviation is 0.98°.The method in this study and the connecting rod wheel angle sensor were used on paddy farm machinery.The wheel angle measurement deviation of the proposed method and the connecting rod wheel angle sensor was not more than 1°,which is relatively small.It has good stability,speed adaptability,and dynamic responsiveness that meets the accuracy requirements of steering wheel angle measurement for paddy field agricultural machinery unmanned driving and can be used instead of connecting rod angle sensors for unmanned agricultural machinery.

Optimization design of the angle detecting system used in the fast steering mirror

In this paper, in order to design a fast steering mirror(FSM) with large deflection angle and high linearity, a deflection angle detecting system(DADS) using quadrant detector(QD) is developed. And the mathematical model describing DADS is established by analyzing the principle of position detecting and error characteristics of QD. Based on this mathematical model, the variation tendencies of deflection angle and linearity of FSM are simulated. Then, by changing the parameters of the DADS, the optimization of deflection angle and linearity of FSM is demonstrated. Finally, a QD-based FSM is designed based on this method, which achieves ±2° deflection angle and 0.72% and 0.68% linearity along x and y axis, respectively. Moreover, this method will be beneficial to the design of large deflection angle and high linearity FSM.

Reactive Navigation of Underwater Mobile Robot Using ANFIS Approach in a Manifold Manner

Learning and self-adaptation ability is highly required to be integrated in path planning algorithm for underwater robot during navigation through an unspecified underwater environment. High frequency oscillations during underwater motion are responsible for nonlinearities in dynamic behavior of underwater robot as well as uncertainties in hydrodynamic coefficients. Reactive behaviors of underwater robot are designed considering the position and orientation of both target and nearest obstacle from robot s current position. Human like reasoning power and approximation based learning skill of neural based adaptive fuzzy inference system（ANFIS）has been found to be effective for underwater multivariable motion control. More than one ANFIS models are used here for achieving goal and obstacle avoidance while avoiding local minima situation in both horizontal and vertical plane of three dimensional workspace.An error gradient approach based on input-output training patterns for learning purpose has been promoted to spawn trajectory of underwater robot optimizing path length as well as time taken. The simulation and experimental results endorse sturdiness and viability of the proposed method in comparison with other navigational methodologies to negotiate with hectic conditions during motion of underwater mobile robot.