Use of Advance Driver Assistance System Sensors for Human Detection and Work Machine Odometry
Siddiqui, Shadman Razzaq
Permanent address of the item is
This master thesis covers two major topics, the first is the use of Advance driver assistance system (ADAS) sensors for human detection, and second is the use of ADAS sensors for the odometry estimation of the mobile work machine. Solid-state Lidar and Automotive Radar sensors are used as the ADAS sensors. Real-time Simulink models are created for both the sensors. The data is collected from the sensors by connecting the sensors with the XPC target via CAN communication. Later the data is later sent to Robot operating system (ROS) for visualization. The testing of the Solid-state Lidar and Automotive Radar sensors has been performed in different conditions and scenarios, it isn’t limited to human detection only. Detection of cars, machines, building, fence and other multiple objects have also been tested. Moreover, the two major cases for the testing of the sensors were the static case and the dynamic case. For the static case, both the sensors were mounted on a stationary rack and the moving/stationary objects were detected by the sensors. For the dynamic case, both the sensors were mounted on the GIM mobile machine, and the machine was driven around for the sensors to detect an object in the environment. The results are promising, and it is concluded that the sensors can be used for the human detection and for some other applications as well. Furthermore, this research presents an algorithm used to estimate the complete odometry/ ego-motion of the mobile work machine. For this purpose, we are using an automotive radar sensor. Using this sensor and a gyroscope, we seek a complete odometry of the GIM mobile machine, which includes 2-components of linear speed (forward and side slip) and a single component of angular speed. Kinematic equations are calculated having the constraints of vehicle motion and stationary points in the environment. Radial velocity and the azimuth angle of the objects detected are the major components of the kinematic equations provided by the automotive radar sensor. A stationary environment is a compulsory clause in accurate estimation of radar odometry. Assuming the points detected by the automotive radar sensor are stationary, it is then possible to calculate all the three unknown components of speed. However, it is impossible to calculate all the three components using a single radar sensor, because the latter system of equation becomes singular. Literature suggests use of multiple radar sensors, however, in this research, a vertical gyroscope is used to overcome this singularity. GIM mobile machine having a single automotive radar sensor and a vertical gyroscope is used for the experimentation. The results have been compared with the algorithm presented in  as well as the wheel odometry of the GIM mobile machine. Furthermore, the results have also been tested with complete navigation solution (GNSS included) as a reference path.