Using Cost-Benefit Analysis to Select Autonomous Vehicle Sensors

Carlos Asuncion (University of California, Berkeley), Rachel Powers (University of Arizona)

In this project, a cost-benefit analysis of LIDAR sensors that are available for use in autonomous vehicles is performed. This will be done by first determining the lowest level of data required for the vehicle to perceive important information from its surroundings. The vehicle will use information from its sensors to fill in an occupancy grid. After finding the minimum performance required for safe operation, research will be done into LIDAR sensors on the market that meet these standards. Data from the Velodyne 64e will be collected and manipulated to simulate some of the lower end LIDAR devices by adding noise or restricting the angle of inclination. The data will be analyzed to see if each sensors is able to correctly detect obstacles and plan a route to a pre-programmed destination. The analysis will be completed under normal conditions and with simulated weather conditions. When the simulations are completed, the vehicle will be tested to see if it can safely find a path with limited LIDAR performance. The sensor selection project will take cost into consideration when determining which sensor best fits the requirements. Finding the best cost-performance balance will allow autonomous vehicle technology to be available to more people. People who are interested in developing systems for autonomous vehicles will be confident that they can operate their vehicle safely with a particular LIDAR configuration.