The behavior of an unmanned Ego Vehicle-such as an Autonomous Vehicle (AV)-in a scenario where a field experiment is too risky or time-consuming to conduct is better studied in simulation; e.g. when a mobile vehicle cuts in front of the ego vehicle, perhaps colliding with it. In simulation, we may synthesize detections of the mobile target vehicle to determine the response of the ego motion planning stack. When the field test is too laborious to conduct-such as one involving two ego vehicles operating synchronously in each other’s vicinity-we may use the first ego vehicle’s perception output to simulate those of the second.In this paper, we pose more precisely and solve the above problem pertaining to the synthesis of target detections from an ego radar. We focus primarily on the radar because as a position and velocity sensor, it allows a wider variety of synthesis problems to be solved compared to position sensors such as camera and lidar; nonetheless, much of our approach is readily applicable to these position sensors as well. To facilitate efficient simulation, we develop a simple geometrical model of the ego radar’s Field of View (FOV) that neglects the physical details of radar wave propagation. We illustrate the results of our approach using AVs as ego vehicles; nonetheless, our methods are agnostic to the robot motion model, as we demonstrate in a synchronized motion simulation between an AV and an Unmanned Aerial Vehicle (UAV/drone). We conclude with a critique of our approach and how future work would address its limitations.
Synthesizing Radar Detections of Target Objects for Unmanned Vehicle Behavioral Simulation
2022-06-01
4068338 byte
Conference paper
Electronic Resource
English
Vehicle localization based on radar detections in garages
European Patent Office | 2023
|Vehicle Localization Based on Radar Detections in Garages
European Patent Office | 2023
|Tracking Vehicles Using Radar Detections
British Library Conference Proceedings | 2007
|Tracking vehicles using radar detections
IEEE | 2007
|