Graduate Student Develops a Process for Testing Automated Vehicle Safety
Mayur Patil has developed multiple test scenarios for a variety of Automated Driver Systems (ADS) including lane departure warnings, lane keep assists, autonomous emergency braking and adaptive cruise control. For each of these systems Patil must apply the proper test scenario and automate that testing process.
Patil is a graduate student working on his master’s degree at Ohio State’s SIM Center at The Ohio State University. As part of his master’s thesis Patil is working on a project sponsored by the Transportation Research Center where he is developing test scenarios for automated vehicle safety systems.
As part of Patil’s project, customers will approach him with certain specifications for the testing of their ADS. Patil first analyzes those requirements and develops the test cases or chooses from a pool of pre-developed test cases. These test cases are developed by defining the Operation Design Domains (ODDs), the object detection responses, the environmental hazards, these combinations will give you a list of test scenarios.
“When a customer comes up with a particular ADS, for example, adaptive cruise control, we analyze the requirements, develop test scenarios and define the ODDs,” Patil said. “For example, that it can operate in conditions such as heavy rains or landscapes with steep inclines. After defining those, you define the object event detection response, which is what the vehicle will do upon detecting any kind of obstacle. After this is designed you develop a test scenario, then when you have all the test scenarios in this process, you select those scenarios that the customer wants.”
The functional safety of the ADS is also being tested. The developed scenarios ensure that all of the functions do not cause any hazard that may cause the loss of life. A fault is injected into the software as it preforms the test scenario- by purposefully injecting a fault into the algorithm, the ADS is being tested to see if it’s backup system, or fail-safe will work.
After a scenario is selected, it is simulated on Software in the Loop, then a risk estimation of the failure of the system is performed. Risk estimation investigates the failure of the particular system and how many times that particular ADS will fail. Using a top down fault tree analysis approach each layer is examined to discover the root causes of the failures.
“In Software in the Loop we develop a model and a vehicle package. There is a scenario blockset in that model too. You input all your vehicle velocities, longitudinal velocities, lateral velocities, everything the scenario reader needs. The blockset gives the lane markings and what actors are involved. You send it to all the sensor models, and feed it back to the vehicle model and you actuate a proper action”.
Patil is automating this entire testing system, thus reducing the overhead processes and workloads. The clients are given a framework, or procedure to test their ADS.
By developing this automated simulated system of testing automated vehicle safety, Patil’s research could contribute in cutting the costs of vehicle testing drastically. To ensure an automated vehicle is road ready it must log hundreds of thousands of miles- by simulating this process, both the cost and the time can be significantly reduced.
Written by Muhammed Al Refai, CAR Marketing and Communications Intern