Design and Implementation of Multi-Sensor Data Fusion Platform for Fuel Saving Lane-Changing Assistance

• Main Authors: Po-ChuanChou, 周柏全
• Other Authors: Chung-Ping Young
• Format: Others
• Language: en_US
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/n2w94u

碩士 === 國立成功大學 === 資訊工程學系 === 104 === The issues of intelligent vehicles are now widely discussed. Increasing road traffic safety and at the same time reducing fuel consumption is one of the most challenging future tasks. In order to improve road traffic safety, the in-vehicle system (On-Board Unit, OBU) must have the ability to perceive complex environment information and detect potential threats on the road. For that reason, we create a multi-sensor data fusion system to process environment information. According to the paper in [20], with maximized traffic throughput, the global fuel consumption will be improved in normal traffic conditions. We develop a lane-changing mechanism that can make a lane-changing decision if there is a forward vehicle blocked the way we go forward, and then we could not pass through the traffic light in front. The purpose of lane-changing mechanism is to maximize traffic throughput for achieving global fuel saving by decreasing vehicles jammed on the road. The data sources include IMU (Inertial Measurement Unit), ultrasound sensor, lidar, speed sensor, camera, RSU (Road Side Unit), OBD-II (On-Board Diagnostics) and GPS. We use Kalman filter to stabilize significant parameters for adjusting safe distance while performing lane-change. In addition to perceive surrounding of ego vehicle, the traffic situation assessment of current road is also an important task. So we develop a novel traffic situation assessment method based on Bayesian classification and a decision algorithm to decide whether the occasion is appropriate to change lane or not. Finally, we built the fusion system with user interface, which can perform the recommendation of lane-change to the driver. In the experiment results, we analyzed and discussed about the performance of specific parameters estimation and traffic situation assessment. Moreover, the system’s response time (from the procedure of accessing raw data to the procedure of showing the lane-changing result) is approximately 0.3 second that measure up to our expectation.