Modeling Perception Errors towards Robust Decision Making in Autonomous Vehicles

Authors: Andrea Piazzoni, Jim Cherian, Martin Slavik, Justin Dauwels

IJCAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this paper, we propose a simulation-based methodology towards answering this question. At the same time, we show how to analyze the impact of different kinds of sensing and perception errors on the behavior of the autonomous system.
Researcher Affiliation Academia 1ERI@N, Interdisciplinary Graduate School, Nanyang Technological University, Singapore 2Centre of Excellence for Testing & Research of AVs, Nanyang Technological University, Singapore 3School of Electrical and Electronic Engineering, Nanyang Technological University, Singapore
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code No The paper mentions using open-source tools like LGSVL simulator [LGSVL, 2019] and Apollo 3.5 [Fan et al., 2018], but it does not provide concrete access to the source code for its own developed methodology (Perception Error Model or PEM).
Open Datasets Yes Scenarios used in experiments: (a) a representative illustration, (b,c) instances from the nu Scenes dataset [Caesar et al., 2019].
Dataset Splits No The paper describes simulation scenarios and parameters for injecting errors but does not specify traditional dataset splits (training, validation, test sets) as it focuses on evaluating system behavior rather than training a machine learning model from scratch with specific data splits.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, memory) used for running its experiments.
Software Dependencies No The paper mentions "LGSVL simulator [LGSVL, 2019]" and "Apollo 3.5 [Fan et al., 2018]" as software tools used. While Apollo has a version (3.5), the paper does not list specific version numbers for other ancillary software libraries or dependencies that would be needed for replication.
Experiment Setup Yes We developed python scripts to implement different scenarios, to automate the tests, configure the simulation environment and the actors in a deterministic manner, and to log the results. [...] Implementation TC1-3. We model the False negative errors by means of Markov chains with two states. We tested different values of the parameters steady state probability [0.0, 1.0] and mean sojourn time (0.0s, 10s]. [...] Implementation TC4, TC5a. Gaussian White Noise with varying standard deviation σ, applied to the relative position of w w.r.t. the AV, in polar coordinates: multiplicative noise on radius d as σd [0%, 12%]; additive noise on azimuth θ as σθ [0 , 1.5 ]. [...] Implementation TC5b. Perfect detection at each frame, but with a tracking loss probability ptl [0, 1] for the previously detected obstacles. [...] These sets are executed multiple times (at least 30 runs per set), to account for the uncertainty introduced by the randomness involved in our PEMs.