Repository navigation
Add LiDAR noise and dropout robustness benchmark (#279) - #306
Merged
ashishsinghbora merged 1 commit intoNov 10, 2026
Merged
ashishsinghbora merged 1 commit into
ashishsinghbora merged 1 commit into
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds a LiDAR robustness benchmark for the 3D drone navigation environment, evaluating a fixed PPO policy under five standardized sensor conditions.
The benchmark uses 30 matched evaluation seeds per condition and deterministic policy actions, without retraining between conditions.
Changes
adaptive-rl benchmark lidarCLI command.Validation
Example results
Using a nominal PPO checkpoint trained for 50,000 timesteps, severe sensor degradation increased the observed collision rate from 26.7% to 46.7%. Paired episode comparisons identified seven newly occurring collisions and one avoided collision under the severe condition.
Performance was not monotonically worse across all sensor conditions. Moderate dropout and combined perturbations produced higher success rates than ideal sensing. These findings are based on one trained policy and 30 evaluation seeds per condition, and should be interpreted as descriptive rather than statistically conclusive.
The full experimental setup, results, and limitations are documented in
docs/BENCHMARKS.md.Related to #279.