You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
M3: Motion Distortion Correction and Timing Robustness Evaluation #747
Integrate LiDAR motion distortion correction into the Awkernel localization pipeline and evaluate its accuracy and timing behavior under delayed, missing, and overloaded input conditions.
Scope:
Motion distortion correction using per-point timestamps and vehicle motion estimates
Integration before downsampling stages that discard required per-point timing information
Validation of timestamp alignment, coordinate transforms, and motion-data availability
Controlled injection of input delays, timestamp offsets, dropped messages, and computational overload
Explicit policies for stale or missing inputs, buffer limits, and processing overruns
Measurement of localization accuracy, end-to-end latency, input data age, deadline misses, and dropped-data rates
Completion criteria:
Motion distortion correction operates within the integrated preprocessing, NDT, and EKF pipeline.
Corrected point clouds and localization outputs satisfy predefined tolerances against reference results under nominal conditions.
Timing and abnormal-input scenarios are reproducible, with documented expected behavior and acceptance criteria.
Stale inputs, missing data, and processing overruns are detected and handled according to the defined policies.
Accuracy and timing trade-offs are reported with datasets, configurations, and reproduction steps.
Out of scope:
Live sensor fault recovery, vehicle-level safety certification, closed-loop vehicle control, and formal worst-case execution-time guarantees.
Activity