Investigation of collision estimation with vehicle and pedestrian using CARLA simulation software

Beg, Mohammad Sojon; Ismail, Muhammad Yusri · 2024 · Crossref

DOI: 10.15282/jmes.18.1.2024.11.0786

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the critical need for effective object detection systems in automotive safety, particularly in complex driving environments such as T-junctions, cross-junctions, and roundabouts. Motivated by the high costs and safety risks associated with real-world testing of Advanced Driver Assistance Systems (ADAS), the authors utilize the CARLA open-source simulator to investigate collision estimation and vehicle-pedestrian interactions. The research aims to evaluate how key factors, including vehicle speed, distance, and traffic density, influence accident probability, thereby providing a cost-effective method for developing context-aware detection systems. The methodology employs the CARLA simulation platform, configured with Python scripting to capture image data and simulate various traffic scenarios. The system differentiates between vehicles and non-vehicle objects, using a visual indicator—a red frame—to denote unsafe distances or imminent collision risks. The study applies the law of conservation of momentum to analyze collision dynamics, calculating momentum ($P=mv$) based on vehicle mass and velocity. Simulations were conducted using specific vehicle models, including a Ford Mustang (1109 kg) and a Dodge Charger (2014 kg), under varying speed conditions ranging from 15 km/h to 105 km/h. Additionally, pedestrian crossing scenarios were simulated to assess braking distances and collision outcomes at both compliant (30 km/h) and excessive (57 km/h) speeds. The results demonstrate a direct and significant correlation between vehicle speed and collision momentum. In vehicle-to-vehicle simulations, increasing speed from 15 km/h to 105 km/h raised momentum from 8,038 kg·m/s to 57,403 kg·m/s, with the rate of increase becoming more pronounced at higher velocities. Specifically, exceeding the 30 km/h speed limit resulted in a momentum of 30,599 kg·m/s upon collision, compared to lower values at compliant speeds. In pedestrian scenarios, driving at 57 km/h led to late braking and collisions, whereas adhering to the 30 km/h limit allowed for safe stopping distances. The object detection system successfully identified hazards, visually marking unsafe proximity with red frames, confirming its utility in real-time hazard recognition. The study concludes that simulation-based analysis is a viable and efficient alternative to real-world testing for developing ADAS. It highlights that exceeding speed limits significantly increases stopping distances and collision momentum, thereby elevating accident risks. The findings underscore the importance of integrating accurate object detection with momentum-based safety assessments to enhance autonomous driving systems. Future research is recommended to explore the impact of adverse weather conditions, such as rain and haze, on sensor performance, particularly LiDAR, to further refine detection accuracy in diverse atmospheric situations.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• situational awareness