WO2024054178A1

WO2024054178A1 - Method of determining driving conditions for total emission optimisation in autonomous vehicles

Info

Publication number: WO2024054178A1
Application number: PCT/TR2023/050823
Authority: WO
Inventors: Engin Ozatay; Orhan Behic Alankus
Original assignee: Istanbul Okan Universitesi
Priority date: 2022-09-07
Filing date: 2023-08-17
Publication date: 2024-03-14

Abstract

The invention relates to a method of determining the driving conditions for emission optimisation in autonomous vehicles that calculate the required speed and acceleration values for the total optimum emission according to the road conditions.

Description

METHOD OF DETERMINING DRIVING CONDITIONS FOR TOTAL EMISSION OPTIMISATION IN AUTONOMOUS VEHICLES

Technical field of the invention

The invention relates to a method of determining the driving conditions for emission optimisation in autonomous vehicles that calculate the required speed and acceleration values for the total optimum emission including particles from brakes and tires according to the road and traffic conditions.

State of the Art

Emission from vehicles is critical for the atmosphere. In addition to the emissions from the engine, the particles emitted from the brakes and tires significantly affect human health. Fuel consumption also causes CO2 emissions. Emissions of harmful gases and particles into the atmosphere cause problems such as global warming and climate change and significantly affect human health. The way of driving has an important effect on the formation of these emissions.

Studies have shown that driving behaviour has a strong impact on traffic safety, fuel or energy consumption, and gas emissions. Automatic recognition of driving behaviour plays an important role in improving overall safety for a driver and in reducing vehicle fuel, energy consumption, and gas emissions from brakes and tires. In addition, monitoring of driving behaviour meets the needs of multiple markets (car insurance, fleet management and fuel consumption optimisation).

Studies conducted in the state of the art propose various systems for recognising and classifying driving behaviours. Among these, smartphones are often used for driving behaviour analysis and data collection. Smartphones have several advantages compared to other data collection devices such as cameras and telematics boxes. First, smartphone-based solutions are scalable, upgradeable, and inexpensive. Second, smartphones can perform online assessments and provide instant driver feedback. Finally, smartphones can offer a shortcut to new technologies due to their short replacement and development cycles. In autonomous vehicles, vehicles need to make many decisions such as understanding the traffic and determining the right route according to the traffic situation. After determining the route, it is important to determine the ideal speed profile for that route. Today, the speed profile is usually calculated according to the allowed speed limits and the traffic situation.

The invention that is the subject of the application numbered "CN110863888A" in the state of the art provides systems and methods for reducing cold start emissions for autonomous vehicles. It relates to reducing emissions and improving fuel economy for vehicles participating in car sharing models. It cannot produce a speed and acceleration profile for autonomous vehicles and only improves cold running.

The invention that is the subject of the application numbered "CN108583165A" in the state of the art relates to a tire structure that detects the amount of wear itself for autonomously operating vehicles. The tire structure, which self-detects the amount of wear, includes a hub, a rubber tire and a control system. In the invention, there is no system that optimises tire emissions and gives speed and acceleration values in integration with other emissions.

In the studies in the state of the art, the speed of the vehicle is determined after the route determination in autonomous vehicles, the vehicle brake and gas data are given to the vehicle, and the vehicle is provided to go on a certain route at a certain speed. Speed is determined according to traffic conditions and speed limits on the road. In determining the speed, reaching the destination in the minimum time or with the minimum fuel consumption is included in various scientific studies. However, no speed and acceleration study has been conducted to optimise both braking, tire and exhaust emissions, and the application of this study to autonomous vehicles has not been found. For this reason, there is a need for a system that aims to optimise driving style, speed, acceleration and to reduce all emissions, including brakes and tires.

The aim of the invention

The invention relates to a method of determining the driving conditions for emission optimisation in autonomous vehicles that calculate the speed and acceleration values required for the optimisation of emissions originating from fuel consumption, engine, brakes and tires according to road conditions.

The most important aim of the invention is to calculate the speed and acceleration values required for the optimum total emission with artificial intelligence according to the road and traffic conditions.

Another aim of the invention is to provide a speed profile for autonomous vehicles by determining the best score for overall emission optimisation for the driving style.

Another aim of the invention is to produce different speed profile for optimum total emission according to traffic and road conditions.

Another aim of the invention is to determine the fuel consumption, air quality and exhaust gases affecting human health, and particles originating from brakes and tires, in a way that reaches the best values for the speed profile detection of autonomous vehicles. Thus, the speed profile can be designed to optimise both fuel consumption, human health and air pollution.

The structural and characteristic features of the invention and all its advantages will be understood more clearly by the figures given below and the detailed description written with reference to these figures. For this reason, the assessment should be made by taking these figures and detailed explanation into consideration.

Description of the invention

The invention relates to a method of determining the driving conditions for emission optimisation in autonomous vehicles that calculate the required speed and acceleration values for the total optimum total emission according to the road conditions.

By using smartphone sensors including GPS and accelerometer, data collection, driving events such as acceleration, deceleration, cornering and acceleration are detected. An acceleration event is a driving event in which the driver accelerates and exceeds the predefined limit for a specified period of time. Similarly, deceleration is an event in which the driver decelerates, either by braking or simply by drawing in, under the influence of the road topology, which can be a straight road or a sloping road. Cornering is another event where the driver turns the steering wheel, and the event is detected by lateral acceleration values collected from the accelerometer sensor. The GPS on the smartphone sensor detects the vehicle's latitude, longitude, speed and direction information. Data can also be obtained from different sensors in the vehicle via the vehicle's computer system.

First of all, sensor data is collected from the on-board computer and/or mobile phone. Then, the calibration parameters are determined in such a way that the z direction of the value calibrated by the processor based on the accelerometer values received from the sensor indicates the centre of the earth, the x direction the longitudinal direction of the vehicle and the y direction the lateral movement of the vehicle. After the calibration parameters are determined by the processor, the accelerometer signals are combined with the GPS data containing latitude, longitude, speed and direction information. The sensor data is analysed in real time by the processor and based on the excitation of the signals, the areas where events such as acceleration, deceleration, and cornering occur are determined. Then, the process continues with the steps of feature extraction and feature selection. These steps provide input for the random forest algorithm, which eventually categorizes the event as acceleration, deceleration, cornering, and acceleration driving behaviours.

The database stores the velocity profile information against the exhaust gases in it. Using the speed profile information in the database, brake wear particles and tire wear particles are calculated by the processor.

Brake wear is calculated by Equation-1 ;

Equation-1 In Equation 1 , M is the mass of the vehicle, Vi is the speed of the vehicle when it starts braking, V2 is the speed of the vehicle when it stops braking, p is the density of the brake pad or disc, <p is the coefficient for the temperature of the brake pad, N is the number of brake gears of each vehicle, k is the fracture coefficient, FN is the normal force of the contact surface, K is the specific wear rate for the material component, and L is the relative sliding distance.

<t> can be taken as 1 .8 since the temperature of the brake pad should not exceed 250 degrees unless repeated hard braking is applied, p brake lining density is taken as 1 .89 g/cm³, k friction coefficient as 0.35, conventional brake lining wear rate as 3.80 *10-6 g/m and N=4.

Tire wear particles are calculated by Equation-2;

Equation-2

Friction force (w) per unit contact area is calculated by Equation-3;

Equation-3 mT is the mass loss, <p is the transverse reduction coefficient due to the tire pattern, w is the friction force per unit contact area, k1 and k2 are two constants characterising the wear behaviour of the rubber compound at a given temperature and on a given abrasive surface, B is the contact width between the tire and the ground, D is the distance of the vehicle driven, N is the number of the vehicle's tires, L is the contact length between the tire and the ground, and P(t) is the energy consumption. The following are taken as shown: k1 = 4E-14, k2 = 1.9, cp=6.

P(t) is calculated with the help of Equation-4. q M C_r dv) [ v(t) + c₂] + g M G(t) + (1 + A) M —

1000 dt)

Equation-4 In Equation-5, Ch refers to height, g to gravity, p to air density, G(t) to slope, M to vehicle weight, Af to vehicle front area, CD to aerodynamic drag coefficient, Cr to rolling resistance factor, Ci and C2 to rolling resistance factor, r|d to powertrain efficiency, and A to Roti masses.

In parallel with the developed R&D studies, different formulas can be used for brake and tire wear.

Then, in the feature extraction step by the processor, with preprocessing, all signals are first linearly interpolated to a fixed sampling rate. The sensor data is then normalized to [-1 ,1 ] using the maximum and minimum values independently. A sliding window is then implemented to split the continuous sensor data into windowed data without overlapping. There are two types of labels in the dataset. The dataset used to train the algorithm only provides the initial timestamp of each class and the corresponding label.

In the feature extraction step, statistical values and automatically extracted features are calculated. Statistical properties are calculated for each window in statistical values, including mean, variance, standard deviation, minimum, maximum, range, and slope between maximum and minimum. Here, the acceleration value a(t) is calculated in Equation-5 by using the acceleration components ax(t), ay(t) and az(t) in the three axes taken from the phone or vehicle accelerometer. In addition, the jolt (j(t)) is also calculated using the accelerometer and the time interval of interest and is used as a component controlling passenger comfort.

Equation-5

In automatically extracted features, both PCA (principal component analysis) and SSAE (Stacked Sparse Autoencoder) are applied to pre-processed data. In addition to the raw data, the acceleration magnitude is included for training both algorithms. For the dataset used to train the algorithm, when C>100, the size of each layer is C^C/2^30^20^T. When C<100, the size of each layer is C^30^20^T. Where C is the size of the input data, T is the size of the resulting feature set. In addition, the drop-out technique is applied on the training data to avoid complex pair adaptations and to avoid repeated extraction of the same features.

Random forest (RF) is a tree-based artificial intelligence algorithm running on the processor that creates a certain number of classification trees without pruning. The average output of the random forest algorithm gives us the most probable driving event. The input of the random forest algorithm is the driving data, and the output is the driving values with negative emissions.

There is emission impact information on the database. The effect of tire and brake particle emissions calculated by the processor is calculated using the database. The processor calculates the maximum speed profile by obtaining the multi-purpose function, cost, fuel consumption, brake, and tire operations, with the optimisation method for the speed limit and the best possible time planning in traffic. In the preferred embodiment of the invention, the processor provides at least 10 different speed profiles to be produced according to traffic and road conditions. Calculated speed profiles are scored by the processor according to the best overall emission values. The processor determines the speed profile by selecting the highest scoring speed profile.

Method of determining driving conditions for total emission optimisation in autonomous vehicles comprises the process steps of

- Collecting sensor data using smartphone sensors including GPS and accelerometer or information on the on-board computer,

- Determining the calibration parameters of the calibrated value based on the accelerometer values received from the sensor, so that the z direction indicates the centre of the earth, the x direction the longitudinal direction of the vehicle and the y direction the lateral movement of the vehicle,

- Combining accelerometer signals with GPS data including latitude, longitude, speed and direction information after the calibration parameters are determined, - Detection of the areas where acceleration, deceleration, cornering and acceleration occur by the processor based on the stimulation of signals by analysing the sensor data in real time,

- Calculation of brake wear particles and tire wear particles by the processor using the speed profile information in the database,

- Extracting features for driving characteristics by the processor,

- Categorisation of deceleration, cornering and acceleration driving behaviours by the processor using a random forest algorithm trained with ready test data,

- Calculation of the effect of the calculated tire and brake particle emissions by the processor using the emission effect information in the database,

- Calculation of more than one optimum speed profile with speed limit and time optimisation by the processor by obtaining cost, fuel consumption, brake, and tire operations, and

- Selection of the speed profile with the highest score by the processor by scoring the calculated speed profiles according to the total emission values.

In the process step of the categorisation of deceleration, cornering and acceleration driving behaviours by the processor using a random forest algorithm trained with ready test data,

- The artificial intelligence training set can be prepared with existing test and/or simulation data for all different vehicles such as trucks, light commercial vehicles, and passenger cars. The weights of the training set for the artificial intelligence method, which are calculated by considering the fuel consumption, emissions, particles, air quality, and the effects on human health, are taken into consideration.

- The artificial intelligence training set can be prepared according to fuel consumption alone or according to different parameters that may include fuel consumption, exhaust gases, tire, and brake emissions.

In the process step of extracting features for driving characteristics by the processor operation; features such as sudden acceleration, high speed, calm driving are extracted.

Claims

CLAIMS Method for determining driving conditions for total emission optimisation in autonomous vehicles, comprising the process steps of

- Combining accelerometer signals with GPS data including latitude, longitude, speed and direction information after the calibration parameters are determined,

- Detection of the areas where acceleration, deceleration, cornering and acceleration occur by the processor based on the stimulation of signals by analysing the sensor data in real time,

- Extracting features for driving characteristics by the processor,

- Calculation of more than one optimum speed profile with speed limit and time optimisation by the processor by obtaining cost, fuel consumption, brake and tire operations, and

- Selection of the speed profile with the highest score by the processor by scoring the calculated speed profiles according to the total emission values. Method for determining driving conditions for total emission optimisation in autonomous vehicles according to Claim 1 , wherein, in the process step of the categorisation of deceleration, cornering and acceleration driving behaviours by the processor using a random forest algorithm trained with ready test data, the artificial intelligence training set can be prepared according to fuel consumption alone or according to different parameters that may include fuel consumption, exhaust gases, tire and brake emissions. Method for determining driving conditions for total emission optimisation in autonomous vehicles according to Claim 1 , wherein, in the process step of the categorisation of deceleration, cornering and acceleration driving behaviours by the processor using a random forest algorithm trained with ready test data, the artificial intelligence training set can be prepared with existing test and/or simulation data for all different vehicles such as trucks, light commercial vehicles, and passenger cars. Method for determining driving conditions for total emission optimisation in autonomous vehicles according to Claim 1 , wherein, in the process step of extracting features for driving characteristics by the processor operation, features such as sudden acceleration, high speed, calm driving are extracted.