An end-to-end approach to autonomous vehicle control using deep learning

The objective of this work is to develop an autonomous vehicle controller inside Grand Theft Auto V game, used as a simulation environment. It is used an end-to-end approach, in which the model maps directly the inputs from the image of a car hood camera and a sequence of speed values to three driving commands: steering wheel angle, accelerator pedal pressure and brake pedal pressure. The developed model is composed of a convolutional neural network and a recurring neural network. The convolutional network processes the images and the recurrent network processes the speed data. The model learns from data generated by a human driver´s commands. Two interfaces are developed: one for collecting in-game training data and another to verify the performance of the model for the autonomous vehicle control. The results show that the model after training is capable to drive the vehicle as well as a human driver. This proves that a combination of a convolutional network with a recurrent network, using an end-to-end approach, is capable of obtaining a good driving performance even using only images and speed velocity as sensory data.

A Fuzzy Drive Strategy for an Intelligent Vehicle Controller Unit Integrated with Connected Data

In order to improve vehicle control safety in intelligent and connected environments, a fuzzy drive control strategy is proposed. Through the fusion of vehicle driving data, an early warning level model was established, and the fuzzy control method was used to obtain the appropriate torque command under the vehicle condition; torque optimization processing was performed according to the different corresponding vehicle following characteristics. The control strategy was tested and verified on an established platform. Based on the experimental results, compared with the traditional drive strategy in one-way front and rear following scenarios, the vehicle avoided excessive opening and closing of the accelerator pedal when the distance between vehicles was close, maintained the correct distance in the following situation, and had better dynamic response when the distance between vehicles was large, indicating that the proposed drive strategy had a better real-time and security performance.

Development of a Control System and Interface Design Based on an Electric Wheelchair

Recently, personal mobility has been researched and developed to make short-distance travel within the community more comfortable and convenient. However, from the viewpoint of personal mobility, there are problems such as difficulty in picking up items while shopping when operating the joystick for shopping and the inability to use hands freely. Accordingly, because the speed of personal mobility can be controlled by foot stepping like an accelerator pedal, we developed an electric wheelchair system that can control the speed by pedal operation. Furthermore, we developed a control system that considers the ride quality using an electric wheelchair with pedal control. In this study, the proposed method is detailed in three parts. Firstly, to develop the pedal mechanism, a potentiometer was used to detect the angle of the pedal mechanism, and a spring mechanism was designed for return to its original position after the pedal was pushed. Next, we propose a feedback control system that considers the ride quality of the operator. In addition, we integrated the system with a smart device-based robot system to realize the mobility as a service (MaaS). Finally, we present several examples of the system and discuss the applicability of the proposed system.

Fuel Economy Improvement of Urban Buses with Development of an Eco-Drive Scoring Algorithm Using Machine Learning

Eco-drive is a widely used concept. It can improve fuel economy for different driving behaviors such as vehicle acceleration or accelerator pedal operation, deceleration or coasting while slowing down, and gear shift timing difference. The feasibility of improving the fuel economy of urban buses by applying eco-drive was verified by analyzing data from drivers who achieved high fuel efficiencies in urban buses with a high frequency of acceleration/deceleration and frequent operation. The items that were monitored for eco-drive were: rapid take-off/acceleration/deceleration, accelerator pedal gradient, coasting rate, shift indicator violation, average engine speed, over speed, and gear shifting under low-end engine speed. The monitoring method for each monitored item was set up, and an index was produced using driving data. A fuel economy prediction model was created using machine learning to determine the contribution of each index to the fuel economy. Furthermore, the contribution of each monitoring item was analyzed using the prediction model explainer. Accordingly, points (defined as the eco-drive score) were allocated for each monitoring item. It was verified that this score can represent the eco-drive characteristics based on the relationship between the score and fuel economy. In addition, it resulted in an average annual fuel economy improvement of 12.1%.

Kinematic and Electrophysiological Characteristics of Pedal Operation by Elderly Drivers during Emergency Braking

Age-related decline in lower limb motor control may cause errors in pedal operation when driving a car. This study aimed to clarify the kinematics and electrophysiological characteristics of the pedal-switching operation associated with emergency braking in the case of elderly drivers. The participants in this study consisted of 11 young drivers and 10 elderly drivers. An experimental pedal was used, and the muscle activity and kinematic data during braking action were analyzed using the light from a light-emitting diode installed in the front as a trigger. The results showed that elderly drivers took the same time from viewing the visual stimulus to releasing the accelerator pedal as younger drivers, but took longer to switch to the brake pedal. The elderly drivers had higher soleus muscle activity throughout the process, from accelerator release to brake contact; furthermore, the rectus femoris activity was delayed, and the simultaneous activity between the rectus femoris and biceps femoris was low. Furthermore, elderly drivers tended to have low hip adduction velocity and tended to switch pedals by hip internal rotation. Thus, the alteration in joint movements and muscle activity of elderly drivers can reduce their pedal operability and may be related to the occurrence of pedal errors.

MATHEMATICAL EXPANSION OF SPECIAL THEORY OF RELATIVITY ONTO ACCELERATIONS

The special theory of relativity (STR) is operationally expanded onto orthogonal accelerations: normal and binormal that complement the instantaneous tangential speed and thus can be structurally extended into operationally complete 4D spacetime without defying the STR. Thus the former classic Lorentz factor, which defines proper time differential can be expanded onto within a trihedron moving in the Frenet frame (T,N,B). Since the tangential speed which was formerly assumed as being always constant, expands onto effective normal and binormal speeds ensuing from the normal and binormal accelerations, the expanded formula conforms to the former Lorentz factor. The obvious though previously overlooked fact that in order to change an initial speed one must apply accelerations (or decelerations, which are reverse accelerations), made the Einstein’s STR incomplete for it did not apply to nongravitational selfpropelled motion. Like a toy car lacking accelerator pedal, the STR could drive nowhere. Yet some scientists were teaching for over 115 years that the incomplete STR is just fine by pretending that gravity should take care of the absent accelerator. But gravity could not drive cars along even surface of earth. Gravity could only pull the car down along with the physics that peddled the nonsense while suppressing attempts at its rectification. The expanded formula neither defies the STR nor the general theory of relativity (GTR) which is just radial theory of gravitation. In fact, the expanded formula complements the STR and thus it supplements the GTR too. The famous Hafele-Keating experiments virtually confirmed the validity of the expanded formula proposed here.

A Sensaptic ADAS Device Using Shape Memory Alloy Wires: Design and Control

This paper presents an active accelerator pedal system based on an integrated sensor and actuator using shape memory alloy (SMA) for speed control and to create haptics in the accelerator pedal. A device named sensaptics is developed with a pair of bi-functional SMA wires instrumented in a synergistic configuration function as an active sensor for positioning the accelerator pedal (pedal position sensing) to control the vehicle speed through electronic throttle and as a variable impedance actuator to generate active force (haptic) feedback to the driver. The reaction force emanated from the pedal alerts the driver and takes appropriate control action by slowing down the vehicle, in harmony with the road’s condition. The design is developed as a proof-of-concept device and is tested and evaluated in a real-time common rail diesel system for rail pressure regulation and over speeding tests, and the responses and performances are found to be promising.

Sound Effects on Physiological State and Behavior of Drivers in a Highway Tunnel

Driving behavior in a highway tunnel could be affected by external environmental factors like light, traffic flow, and acoustic environments, significantly when these factors suddenly change at the moment before and after entering a tunnel. It will cause tremendous physiological pressure on drivers because of the reduction of information and the narrow environment. The risks in driving behavior will increase, making drivers more vulnerable than driving on the regular highways. This research focuses on the usually neglected acoustic environment and its effect on drivers' physiological state and driving behavior. Based on the SIMLAB driving simulation platform of a highway tunnel, 45 drivers participated in the experiment. Five different sound scenarios were tested: original highway tunnel sound and a mix of it with four other sounds (slow music, fast music, voice prompt, and siren, respectively). The subjects' physiological state and driving behavior data were collected through heart rate variability (HRV) and electroencephalography (EEG). Also, vehicle operational data, including vehicle speed, steering wheel angle, brake pedal depth, and accelerator pedal depth, were collected. The results indicated that different sound scenarios in the highway tunnel showed significant differences in vehicle speed (p = 0.000, η2 = 0.167) and steering wheel angle (p = 0.007, η2 = 0.126). At the same time, they had no significant difference in HRV and EEG indicators. According to the results, slow music was the best kind of sound related to driving comfort, while the siren sound produced the strongest driver reaction in terms of mental alertness and stress level. The voice-prompt sound most likely caused driver fatigue and overload, but it was the most effective sound affecting safety. The subjective opinion of the drivers indicated that the best sound scenario for the overall experience was slow music (63%), followed by fast music (21%), original highway tunnel sound environment (13%), and voice-prompt sound (3%). The findings of this study will be valuable in improving acoustic environment quality and driving safety in highway tunnels.

Personalization of Electric Vehicle Accelerating Behavior Based on Motor Torque Adjustment to Improve Individual Driving Satisfaction

As worldwide vehicle CO2 emission regulations have been becoming more stringent, electric vehicles are regarded as one of the main development trends for the future automotive industry. Compared to conventional internal combustion engines, electric vehicles can generate a wider variety of longitudinal behaviors based on their high-performance motors and regenerative braking systems. The longitudinal behavior of a vehicle affects the driver’s driving satisfaction. Notably, each driver has their own driving style and as such demands a different performance for the vehicle. Therefore, personalization studies have been conducted in attempts to reduce the individual driving heterogeneity and thus improve driving satisfaction. In this respect, this paper first investigates a quantitative characterization of individual driving styles and then proposes a personalization algorithm of accelerating behavior of electric vehicles. The quantitative characterization determines the statistical expected value of the personal accelerating features. The accelerating features include physical values that can express acceleration behaviors and display different tendencies depending on the driving style. The quantified features are applied to calculate the correction factors for the target torque of the traction motor controller of electric vehicles. This driver-specific correction provides satisfactory propulsion performance for each driver. The proposed algorithm was validated through simulations. The results show that the proposed motor torque adjustment can reproduce different acceleration behaviors for an identical accelerator pedal input.