Log-Based Simulation Trajectory Interpolation for Autonomous Vehicles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Log-based simulations for autonomous vehicles face issues due to differences in sensor data fields of view and occlusions, leading to objects appearing 'from nowhere' and surprising the simulated vehicle, as the log data does not include absolute 'ground truth' from all perspectives.
Innovation Solution
The method involves analyzing log data to backward or forward interpolate the trajectories of objects by estimating their distance traveled, determining starting or ending locations, and appending trajectories to the log data segment, allowing for realistic simulations by including the location of road user objects before they were observed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If log-based simulations are used for autonomous vehicle testing, then testing efficiency is improved, but simulation realism deteriorates due to objects appearing 'from nowhere' caused by sensor field of view limitations and occlusions
Solution Approach 1:
The system performs preliminary actions by backward interpolating object trajectories to determine starting locations before objects are observed by sensors. This allows objects to be injected into the simulation environment at appropriate earlier positions rather than appearing suddenly, thereby maintaining simulation realism while enabling efficient log-based testing
Solution Approach 2:
The system introduces an intermediary processing layer that takes sensor log data and computed trajectory information as input, then generates enhanced simulation scenarios with properly positioned objects. This intermediary process reconciles the limitations of sensor field of view with the requirements for realistic simulation by computing probable object positions before sensor detection
2Device complexity
If sensor log data is used directly for simulations, then data processing complexity is reduced, but object trajectory accuracy deteriorates due to occlusions and limited field of view
Solution Approach 1:
The system performs preliminary trajectory computation by estimating object starting locations backward from observed positions using speed and time information. This preliminary action fills in trajectory gaps caused by occlusions and field of view limitations without requiring complex real-time processing during simulation execution
Solution Approach 2:
The system enhances the simulation by adding a temporal dimension to object positioning. By computing backward trajectories and injecting objects at earlier time points with calculated positions, the system reconstructs complete object paths that account for occlusion periods, thereby improving trajectory accuracy without proportionally increasing processing complexity
Data Source
AI summary
Aspects of the disclosure relate to improving realism in simulations for testing software for operating a vehicle in an autonomous driving mode. In one instance, an initial observation of a road user object may be identified in a log data segment captured by a perception system of a vehicle. The perception system having one or more sensors. The initial observation includes a point in time and an initial location of the road user object. A distance traveled by the road user object from a start of the log data segment to the point in time may be determined. A starting location for the road user object may be determined using the distance traveled. A trajectory for the road user object may be determined between the starting location and the initial location of the road user object. The trajectory may be appended to the log data segment.