Abstract—Early driverintention prediction plays a sig- nificant role in intelligent vehicles. Drivers exhibit various driving characteristics impairing the performance of con- ventional algorithms using all drivers’ data indiscriminat- ingly. This paper develops a personalized driverintention prediction system at unsignalized T intersections by seam- lessly integrating clustering and classification. Polynomial regression mixture (PRM) clustering and Akaike’s infor- mation criterion are applied to individual drivers trajecto- ries for learning in-depth driving behaviors. Then, various classifiers are evaluated to link low-level vehicle states to high-level driving behaviors. CART classifier with Bayesian optimization excels others in accuracy and computation. The proposed system is validated by a real-world driv- ing dataset. Comparative experimental results indicate that PRM clustering can discover more in-depth driving behav- iors than manually defined maneuver due to its fine ability in accounting for both spatial and temporal information; the proposed framework integrating PRM clustering and CART classification provides promising intention prediction per- formance and is adaptive to different drivers.
5. The mean calculation time of the algorithm evaluated by the test drive is 25.2 ms. The mean calculation time shows real time capability of the algorithm. This means that the algorithm is able to detect all the driving maneuvers in real time and can provide a maneuver warning to the driver on expected time. The results mentioned above provide the solution to all the problems mentioned in section 1.2. The results also successfully satisfies the main research question of this thesis which is “what potential does the driverintention detection algorithm have concerning the detection quality and predicted time horizon with regards to real time requirements”. The detection quality is improved by TPR and FPR values, predicted time horizon is improved by early detection of a driving maneuver and real time requirements are satisfied by the mean calculation time of the algorithm. So by optimization and further development a very good performance for the algorithm is obtained regarding detection quality, predicted time horizon and real time capability. As it can be seen in example table 4.6 that different combinations of results are obtained for stop maneuver and the best combination is found out by the validation metrics. In the similar way, the best combination is found out for all other driving maneuvers. In table 4.6, the stop flag gives 100% TPR, 16.49% FPR and an earliness value of 2.635 seconds. The combination of Stop and Stop 20 flags give 96.11% TPR, 19.48% FPR and an earliness value of 3.095 seconds. For this work, the combination of Stop and Stop 20 flags is selected as the best result because the validation result for this combination is highest.
A systematic transportation system is vital for the development of a country. In Malaysia, the high dependence on private cars is caused by inadequate public transport that does not meet the general needs of commuters. This study was conducted to identify the factors that contributed to drivers’ intention to use electric cars as a greener alternative to fuel- powered vehicles. To explore this issue, Technology Acceptance Model (TAM) was employed to predict acceptance of electric cars based on driverintention. Self-administered questionnaires were distributed to 217 car drivers in the Batu Pahat district. Correlation and regression analyses were performed to determine the relationship between the TAM constructs and the intention to use electric car. Results showed that all contructs were found to be statistically significant. In addition, Perceived Ease of Use was a stronger factor that contributed to drivers’ acceptance to use electric cars compared to Perceived Usefulness. The level of the acceptance was highly positive.
A simplified processing loop of an autonomous car is depicted in Figure 1.2. The process begins with sensing, as this is how the vehicle is able to gather data about the environment it is in. This describes which type of sensors are being used, such as camera, lidar or radar, and these sensors require an appropriate calibration. This raw sensor data is then passed to a perception system, which is able to identify objects and features in the scene, such as lane markings and other vehicles. Parallel to this is a localisation system, which is able to determine the position of the autonomous vehicle on a map. The perception data is then passed to a scene understanding module. This module is able to understand the surrounding scene and predict what the scene will look like in the near future based on context. This includes predictions such as: "A pedestrian is waiting at a crosswalk, therefore they are likely to cross the road." or "A vehicle is slowing down and moving to the outside of the lane as it approaches the intersection, how likely is it that this vehicle will turn into the side street?". This thesis will focus on scene understanding techniques, as it will present methods for vehicle intention and path prediction. The mission planner is then able to incorporate these predictions into a path planning algorithm that allows the autonomous vehicle to reach its intended destination while interacting with traffic. Finally, the mission planner sends commands to the vehicle’s actuators, namely the engine, brakes and steering wheel which allow the vehicle to drive.
force signal acquired, the pedal force sensor has a large gap, so the pedal force is not good to be controlled, and the real-time performance is not good, so the brake pedal displacement signal is usually used as the parameters of the driver's braking intention. Through the bench test, some research scholars take a multi group of brake displacement - brake oil pressure curve with different people used different pedal rate .as the following.
Human driverintention recognition is a difficult task due to it is a highly random signal and unable to measure directly at this moment. It can only be inferred with outer human behaviors or the indirectly measured brain signals. Therefore, in previous research, most of relative research focus on using machine learning methods especially supervised learning to train an intention classifier. However, one of the big challenge of using supervised learning methods is hard to define the true intention label. In contrast, unsupervised learning is suitable for those data that we do not know exactly their labels. Hence, it can be used to cluster the data to the group it most likely belongs to and able to find the intrinsic pattern in the data. In this paper, instead of identifying human intention with supervised learning methods, we use unsupervised learning method to recognize the braking intention. In addition, we compare two different unsupervised learning methods which are K-means and Gaussian Mixture Model to evaluate their different performance on braking intention recognition task. Finally, we also evaluate the contribution of different features to intention
The most advanced ELD systems, like AOBRDs, help automate everyday driver tasks through a line of features like digital inspection reports and truck-grade navigation that takes into account road restrictions and other measures, based on a driver’s vehicle type. Plus, traffic and weigh stations stops are a breeze: HOS data is easily accessible on ELDs, eliminating lengthy processing time. These all-in-one in-cab devices keep drivers on task and managers in the know, empowering companies to stay competitive at every level.
The major task of cost accounting is connected with the allocation of indirect costs to cost objects. To allocate indirect costs (also called as overhead or common costs) to cost objects cost drivers are selected as the cost allocation bases. Selecting the cost drivers is critically important for developing costing methodology. In order to improve the accuracy and credibility of the allocation the most appropriate cost drivers should be selected, and more than one cost driver should be applied. Thus the decision on which and how many cost drivers to use is of critical importance. The number of cost drivers should be optimal, as an excess number of cost drivers could lead to skewed results. The aim of this article is to present the theoretical framework of the cost drivers, including the selection of cost drivers, and to report on the author’s research on cost accounting. The article describes cost drivers, specifies their meaning, gives an overview of their classification, typologies, and also discusses the selection methods, relevant research papers and other issues regarding the cost drivers. The article also provides the first-time review of the study carried out on implementation of cost accounting in universities, including the use of cost drivers. The survey was conducted among different universities all over the world in the period from May to September 2013.
Initially, it was tried to include as many variables (driver, vehicle, environment, roadway, time-related) as possible for the modeling considering the fact that the quality of the modeling could be expected to increase to a certain level once the number of variables increases. Selection of the variables was carried out based on previous studies and on the assumption that a particular variable would affect the severity of ROR crashes. The descriptions of 43 explanatory variables that are considered for the mod- eling are provided along with their statistics in Table 2. All the explanatory variables are binary except SPEED, which is considered as a continuous variable. Binary va- riables take the form of either 0 or 1; for example, if a crash occurs during weekend, the variable WEEKEND has been assigned “1” as its value, otherwise “0” is as- signed to this variable. Three binary logistic regression models were developed by considering crash severity as the response variable and the description of the models are as follows:
monitoring and archiving sent as well as received text messages. Obtaining information from the driver prior to GPS adoption was highly problematic. A Logistics Manager of Firm A argued: “We had to call the driver to find out the position of the vehicle. Many times either the driver did not answer the phone or we could not get through. This was very inconvenient, due to the short reaction time required by the customer. After the conversation with the driver, we put down the information into a form, discussed it with other co-workers and decided which driver is going to go in each direction. Then we called the driver again. If he or she did not answer we sent a text message or called him or her again”. After GPS adoption, dispatchers and other workers used GPS to obtain information required. Only in extreme situations (e.g. when a vehicle lost power, when a dispatcher needed detailed explanation about a certain problem) they actually now had to call the driver.
The proposed model then is implemented into motorcycle taxi dispatch simulation. The simulation is developed based on PHP language. The simulation is not time variant. The world is virtual square area. The length and width are 5 kilometers. The scenario is as follows. First, 60 drivers are created with specific location and idle time. Then 30 passengers arrive and create pickup request consecutively. The passenger arrives with specific pickup location. For driver whom is allocated to the pickup request, the driver then will be deleted from the simulation. The simulation visualization can be seen in Figure 3. Grey circle represents driver that does not get allocation. Red circle represents driver that gets allocation. Red square represents passenger.
The Main Block Diagram OF Project is shown in figure it consist of LPC2148,Line Driver, RF TRANSMITTER module.LPC2148 Connected to line driver and line driver connected to RF TRANSMITTER module and lpc2148 connected to various block these block namely Heart Beat Sensor, Blood Pressure Sensor, Body temperature sensor. This particular sensor gives reading and connected to lpc2148 heart beat sensor gives heart reading (BPM) in 069, 070, 71, 72 etc. takes reading in following format. Blood pressure gives reading in (mm hg), body temperature in farad. All connected to LPC2148 through max 232 through transmitter RF TRANSMITTER module Through Receiver side RF TRANSMITTER Module.
The Generate Distribution Specific Driver Package option provides a simple way to generate a package specific for your distribution. This option uses the policies set by the operating system vendor and allows for maximum compatibility with the distribution. To install the AMD Catalyst proprietary driver using the Generate Distribution Specific Driver Package option, follow these steps:
For development of the questionnaire, the methodology described by Francis et al.  was followed. The behav- iour to be investigated was defined by the elements of Target, Action, Context and Time (TACt). More specif- ically, and following the TACt principle, the ModuleVet behaviour to be investigated was to “contact a veterinar- ian (A) on the same day (t) as a dairy cow with MCM (T) is detected (C)”; for Module2 the behaviour to be investigated was to “start medical treatment (A) on the same day (t) as a dairy cow with MCM (T) is detected (C)”. The first section of the questionnaire addressed the demographics of the farmer, and was followed by a sec- ond section that contained eight case scenarios to meas- ure behavioural intention. The third section contained questions to address the farmer’s attitude, which was measured by three direct questions and 11 behavioural beliefs weighted by their 11 corresponding outcome evaluations. The fourth section addressed the subjective norm, measured by two direct questions and five ques- tions on normative beliefs weighted by five questions assessing the motivations to comply. The fifth section included questions to assess the perceived behavioural control; from this section only one direct question
Drivers are downloadable from the Glyph website at: www.glyphtech.com/support/accessories.php If you choose to download the driver from Silicon Image, go to www.siliconimage.com and go to the support section. Find the SiI3132 PCI Express (1x) to 2 Port SATA 300 and download the cor- rect driver for your version of Windows. Make sure to download the BASE driver, not the SATARAID5 driver.
If the driving licence was issued after 1997 category D1 (minibus) won’t be on the licence. In which case a minibus must to be driven under a “Section 19 Permit” which allows not for profit organisations to transport people providing the driver does it voluntarily.
2.2 Hand-held terminal software architecture Hand-held terminal software architecture Hand-held terminal software is divided into four parts: Boot Loader, Drivers, Kernel, and Applications. Primarily, Boot Loader is used for initializing necessary peripherals and interrupt vector table in processor. Secondly, Drivers, which refer to LCD driver, touch screen driver, NAND Flash driver and wireless module driver are used for supporting communication with Kernel. Thirdly, Kernel chooses real time multitasking kernel for Linux. Finally, Application can be available based on these above mentioned three dispensable parts: Boot Loader, Kernel and Drivers. Software architecture is shown in Figure 2.3