University of Twente Student Theses
Study of Self-Driving Functionality : From SDC to Lane Boundary Detection Based Implementation in Small-Scale Karts
Ashokan, Raj Kumar (2024) Study of Self-Driving Functionality : From SDC to Lane Boundary Detection Based Implementation in Small-Scale Karts.
|
PDF
76MB |
| Abstract: | This paper presents work based on our participation in the Self Driving Challenge(SDC) 2023 edition, aimed at the study of basic autonomous functionality behaviour in cars and discusses the vision-based approaches implemented during the challenge including an unsuccessfully attempted monocular visual odometry-based approach. The runner-up lane boundary marker detection-based method, which made the SDC electric go-kart traverse autonomously a distance of ~1km autonomously, is presented. Due to lack of proper analysis and evaluation of the same during the short stint of the challenge, a small-scale test kart set up was rebuilt including a simple CAN-communication network as in conventional cars. The implementation was adapted to be applicable to challenges pertaining to outdoor small-scale kart environments. A complete pipeline is explored along with the suitable lateral control approaches based on the detected lane boundaries. We also identify a significant discrepancy in a commonly used steer approach claimed to maintain the kart at the center of the lane. Unlike most other related works, real-time testing of the kart on a running track is demonstrated successfully under varying environmental conditions. Comprehensive discussion and extensive evaluation is performed, with the detected lines being evaluated using a newly adapted Mean-Average-Perpendicular-Distance(MAPD) metric with a resulting value of 1.395 pixels. Precision, Recall, and F1-scores of 0.97, 0.99 and 0.97 are achieved for the lane detection . The running of the kart at the center of the lane is evaluated by comparing the derived steer curves with the ideal steer curves generated for the known trajectory and computing a MAE. A mean deviation as low as 2-3 cm from the lane center towards the left is achieved throughout the runs. The algorithm effectively runs at 30FPS during real-time runs. |
| Item Type: | Essay (Master) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Programme: | Electrical Engineering MSc (60353) |
| Link to this item: | https://purl.utwente.nl/essays/104378 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page