the use of lidar technology in autonomous cruise control systems for ...

Gregory JakubiecBrendan MullinixLIDAR IN AUTOMOBILESThe direct application of LiDAR systems that we aremost concerned with is its use in self-driving automobiles.Autonomous cars that utilize LiDAR generally have multipleLiDAR devices onboard, in addition to various otherimportant devices used to sense their surroundings, such asradar, cameras and video systems. They also carry complexcomputer processing systems that analyze the data. It shouldbe noted that LiDAR can’t effectively navigate anautonomous car by itself. It must work in conjunction withvarious other devices, such as cameras and radar systems, aswell as advanced computer hardware and software.There have been multiple specific methods of analyzingand utilizing the LiDAR data that have been experimentedwith, although the majority of them are heavily based on theconcept of grid based processing (which will itself later beexplained in much greater detail), in order to mapenvironments in real time.FIGURE 4LiDAR and Radar IntegrationIn addition to LiDAR, many cars also have radar systemsonboard. Although LiDAR can take measurements acrosslarger ranges and distances, as well as operate much better ininclement weather, radar is able to provide the vehicle withmore detailed information, due to the Doppler Effect.LiDAR technology is our main focus, although radar doeshave its place in autonomous cars as well, and is not to beforgotten. The most efficient autonomous car should utilizeboth [5].When used to operate automobiles, LiDAR andtraditional radar both serve their respective purposes. Inorder for a LiDAR machine to measure velocity, forinstance, it must take different distance measurements atdifferent times, and use them to compute the velocity of theobject being tracked. This leads to a relatively inaccuratevalue of velocity, which in turn leads to an inaccurate valueof acceleration. Radar, however, makes use of the DopplerEffect, which allows systems to measure the disturbances inthe frequency of the return signal, which correspond directlyto velocity of the point being measured. [5]Although LiDAR may not provide as accurate velocityinformation as radar systems, it does operate at much largerranges and fields of view. This is essential to a vehiclemoving through curves and around corners at high speeds[5]. The most efficient autonomous navigation system is onethat utilizes both LiDAR and radar in harmony, covering theweaknesses and exploiting the strengths of both systems.Below is a diagram of one example of an autonomous carsystems utilizing multiple types of sensing technology tonavigate.Diagram of Radar, and Ultrasonic collision detector sensorson a vehicle. [4]Grid Based ProcessingThere have been multiple methods of analyzing andutilizing the LiDAR data that have been experimented with,although the majority of them are heavily based on theconcept of grid based processing.Grid based processing is a general technique used inautonomous cars to map their environments. The basic ideaof this type of data processing is that very small amounts ofdata are each stored in very specific “locations”,corresponding to their actual physical locations, within agigantic grid. This allows the system to quickly andefficiently locate and analyze whatever specific data may berelevant at the current moment, and then update all of it inreal time.The environment around the car is divided up into smallsections of grid within the computing system. The LiDARand radar systems then collect data about each piece of thisgrid, returning it to the analysis software. Using the responsefrom the different systems, a probability is determined for aspecific grid square to be occupied, and assigned to thatspot. With each of these small grid spaces being extremelysmall, a large grid environment is formed [9]. Conclusionscan then be made by the computer regarding the shapes andtextures of objects by analyzing the similarities anddifferences of clusters of grid squares to each other, andsubsequently what types of objects they are most likely tobe. This information is then used to navigate the careffectively throughout this constantly changing grid.4