<p dir="ltr">Most road fatalities are caused by human error. To help mitigate this issue and enhance overall transportation safety, companies are turning to advanced driver assistance systems and autonomous vehicle development. Perception, a key module of these systems, mostly uses light detection and ranging (LiDAR) sensors and enables efficient obstacle detection and environment mapping. Extensive research on the use of LiDAR for autonomous driving has been documented in the literature. Yet still, several researchers and practitioners have advocated continued investigation of LiDAR placement alternatives. To address this research need, this thesis research begins with a comprehensive review of different sensor technologies – camera, radio detection and ranging, global positioning system, and inertial measurement units – and exploring their inherent strengths and limitations. Next, the thesis research developed a methodological multiple criteria framework and implemented it in the context of LiDAR placement optimization. Given the numerous criteria and placement alternatives associated with LiDAR placement, multi-criteria decision analysis (MCDA) was identified as an effective tool for LiDAR placement optimization. MCDA has been applied to some extent in decision-making regarding autonomous vehicle development. However, its application in LiDAR placement optimization remains unexplored. In evaluating the LiDAR placement alternatives, the research first established the placement alternatives and then developed a comprehensive yet diverse set of criteria – point density, blind spot regions, sensor cost, power consumption, sensor redundancy, ease of installation, and aesthetics. The data collection methods included the CARLA simulator, sensor datasheets, and questionnaire surveys. The relative importance among the evaluation criteria was established using weighting techniques such as respondent-assigned weighting, equal weighting, and randomly generated weighting. Then, to standardize the different measurement units, scaling was carried out using value functions developed for each criterion using data from the respondents. Finally, the weighted and scaled criteria measures were amalgamated to obtain the overall evaluation score for each alternative LiDAR placement design. This enabled the ranking of the placement designs and the identification of the best-performing and worst-performing designs. Hence, the optimization method used is the enumeration technique. The findings of this study serve as a reference for future similar efforts that seek to optimize LiDAR placements based on select criteria. Further, it is expected that the thesis’s framework will contribute to an enhanced understanding of the overall impact of LiDAR placement on autonomous vehicles, thus enabling the cost-effective design of their placement and, ultimately, improving AV operational outcomes, including traffic safety.</p>
Identifer | oai:union.ndltd.org:purdue.edu/oai:figshare.com:article/24790644 |
Date | 14 December 2023 |
Creators | Zainab Abidemi Saka (17616717) |
Source Sets | Purdue University |
Detected Language | English |
Type | Text, Thesis |
Rights | CC BY 4.0 |
Relation | https://figshare.com/articles/thesis/LiDAR_PLACEMENT_OPTIMIZATION_USING_A_MULTI-CRITERIA_APPROACH/24790644 |
Page generated in 0.0016 seconds