An Advanced Driver Assistance System For Suzuki Cars
Road accidents are one of the main causes of higher injuries and death rate in the world. According to national highway and traffic safety administration (NHTSA), 36096 people lost their lives in united states of America due to road accidents in 2019 only. Moreover, European safety and transport cou
2025-06-28 16:30:12 - Adil Khan
An Advanced Driver Assistance System For Suzuki Cars
Project Area of Specialization Artificial IntelligenceProject SummaryRoad accidents are one of the main causes of higher injuries and death rate in the world. According to national highway and traffic safety administration (NHTSA), 36096 people lost their lives in united states of America due to road accidents in 2019 only. Moreover, European safety and transport council (ETSC) reported that 22,000 people died in road accidents in 2019 in European Union. In Pakistan, nearly 36,000 people lost their lives in road accidents as stated by Pakistan bureau of statistics (PBA). Among these, 94% are due to human negligence e.g., use of mobile phone, eating, or over speeding while driving on the road. These problems are an indication towards the need of a robust driver assistance system to mitigate these issues.
Due to rapid advancement in artificial intelligence (AI) in the past couple of decades, an advanced driver assistance system (ADAS) is being developed to help human drivers in avoiding these uncalled accidents. The society of Automotive Engineering (SAE) has introduced six autonomy levels, ranging from No-Autonomy (Level 0) to Full-Autonomy (Level 5) for the vehicles. Taking this innovation into account, Mercedes Benz introduced Level-1 and Level-2 autonomy which include autonomous distance-based speed adjustment, emergency braking and steering control in its S-Class vehicles. Moreover, BMW has also introduced ADAS based semi-autonomous features in their X-SAV and I-Series vehicles. However, these driver assistance systems are only compatible with luxury vehicles of their specific brand/model, which only accounts for 5% of the total road traffic. Whereas, 95% of vehicles don’t have such systems, due to their high costs and compatibility issues.
To improve the vehicle safety systems of middle-class vehicles, we are aiming to propose, a low-cost generalized plug-n-play kit-based driver assistance system for vehicles i.e., cars, vans, and buses. The proposed system will assist the driver in robust object detection in adverse conditions and will be able to respond, if the driver fails to take appropriate measures according to the surrounding environment. The proposed system will be capable of object detection and distance measurement which will help the driver in accurate environment analysis. In addition, the proposed system is a Plug-in-Play kit-based solution which can be installed in any kind of vehicle. It is important to mention here that this low-cost solution is first of its kind, which will enhance the road safety features of existing vehicles. Lastly, the proposed system will the first generalized driver assistance system of Pakistan, which will ultimately help in the increase in automotive exports of Pakistan.
Project Objectives.
We are aiming,
- To develop a generalized plug-n-play advanced driver assistance system for vehicles
- To improve the safety features of vehicles at the fraction of a cost
- To minimize the road collisions
- To help in enhancing the safety features of “Made in Pakistan” vehicles i.e., Suzuki.
The proposed system is an embedded system comprising of hardware and software modules. The software module is an object detection system, whereas, hardware module is a Plug-n-Play kit as shown in figure. The development of proposed system will be done through following procedure.
1. Development of Detection System
a. Dataset Collection
Recently, there is no public datasets available for instance segmentation having diverse vehicle classes, which are used in Pakistan. Existing public datasets i.e., Cityscape, & Camvid are constructed over limited driving videos, having quite different classes of road objects (i.e., vehicles, road infra-structure) than Pakistan. Consequently, the performance of object detection system trained over these datasets will be influenced in Pakistani road environment. To overcome these issues, we will construct a diverse dataset over driving videos of Pakistan.
b. Data Pre-processing
After completion of dataset collection, we’ll perform data pre-processing to prepare dataset for annotation. We’ll perform colour correction and frame extraction frame videos to get useful images for annotation.
c. Data Annotation
After data pre-processing, we will start data annotation. To make our object detector more robust, we will perform two types of annotations: (i) 2D Bounding Boxes, and (ii) Panoptic Segmentation.
d. Training Object Detector & Feature Extractor
After completing data annotation, we are aiming to propose custom convolutional neural network-based feature extractor. To avoid over-fitting, initially, we will train the purposed CNN on ImageNet dataset. Later on, it will be fine-tuned on our self-constructed dataset in order to achieve maximum accuracy. Simultaneously, we’ll implement SegNet to perform instance segmentation to achieve desired output.
2. Development of Plug-n-Play Kit
The proposed hardware module will be an on-board unit (Plug-n-Play kit) comprising of high-definition imaging device, microprocessor, an invoking device along with the power source. The imaging device will be placed in-front of the vehicle to capture real-time surrounding environment which will be forwarded to the heavy computing microprocessor for further processing. The proposed object detection system will be installed in the microprocessor which will extract the features from the provided frames in order to detect the objects. Based on the detected objects, system will measure the distance between the bullet and the target object. The proposed system will activate invoking device, if the distance reaches to minimum threshold while driver’s expected response is not executed. It is important to mention here that a small-scale sensing module will be installed in the accelerator, and brake paddles of the vehicle to track the human driver’s response. The graphical representation of the proposed hardware is shown in figure.
- Till recently, a huge gap has been found in-terms of driver assistance systems in 95% of the vehicles by the manufacturers due to high cost of these technologies. Whereas, our proposed system is a low cost generalized Plug-n-Play kit which can be installed in any kind of vehicle without compatibility issues.
- The proposed system will help in reducing the road accidents ratio, which will ultimately reduce the number of fatalities and human injuries.
- Other than ADAS, the proposed system can also be utilized in automating the toll-collection plazas and traffic monitoring systems.
- It will help in increasing the economic growth of automotive industry of Pakistan.
- It will help in forming an industry-academia collaboration, which will pave a way to introduction of automated features in the automotive industry.
The final deliverable of the proposed system will be a Plug-n-Play on-board unit comprising of high-definition imaging device, heavy computing microprocessor, along with invoking device and 12v power source. The proposed object detection system will be installed in the microprocessor. Whereas, developed kit can be installed in the dashboard of the vehicle. While keeping the existing power source of vehicles in mind, the proposed system is also designed to operate on 12v to avoid any extra addition of power source. The suitable imaging device i.e., GoPro Hero 9 is selected for the proposed system to cater with the adverse weather conditions. Further, GPU-enabled computing machine of Nvidia i.e., Jetson Nano will be utilized as microprocessor to execute the proposed object detection system. Whereas, 12v power source will be provided from the vehicle battery to the proposed kit. To track the movement of accelerator and brakes paddles, motion sensing units will be installed in the paddles. The microprocessor will activate the invoking module, if the distance between the source and target vehicle reaches to the minimum threshold, while the movement of the paddles by human driver is not detected.
Final Deliverable of the Project HW/SW integrated systemCore Industry TransportationOther Industries Manufacturing Core Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Industry, Innovation and InfrastructureRequired Resources| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 80000 | |||
| Imaging Device(12.3 megapixels lens, 1.55x1.55pixels size | Equipment | 1 | 12000 | 12000 |
| GPU Enabled Computing Device ( 128core,ARM A57,40 GPIO,4GB DDR4 | Equipment | 1 | 22000 | 22000 |
| Motion Sensor | Equipment | 10 | 450 | 4500 |
| Jumper wires | Miscellaneous | 4 | 250 | 1000 |
| Battery | Equipment | 1 | 9000 | 9000 |
| Buzzer | Equipment | 1 | 500 | 500 |
| PCB | Equipment | 1 | 10000 | 10000 |
| Copper Sheets | Miscellaneous | 1 | 2500 | 2500 |
| DC Adapters | Miscellaneous | 2 | 250 | 500 |
| Vehicle overhead | Miscellaneous | 1 | 5000 | 5000 |
| ferric chloride | Miscellaneous | 1 | 1000 | 1000 |
| ESPs | Equipment | 4 | 3000 | 12000 |