Autonomous golf cars for public trial of mobility-on-demand service

Pendleton, Scott; Uthaicharoenpong, Tawit; Chong, Zhuang Jie; Fu, Guo Ming James; Qin, Baoxing; Liu, Wei; Shen, Xiaotong; Weng, Zhiyong; Kamin, Cody; Ang, Mark Adam; Kuwae, Lucas Tetsuya; Marczuk, Katarzyna; Andersen, Hans; Feng, Mengdan; Butron, Gregory; Chong, Zhuang Zhi; Ang, Marcelo H.; Frazzoli, Emilio; Rus, Daniela · 2015 · OpenAlex-citations

DOI: 10.1109/iros.2015.7353517

archive: archived pipeline: cataloged verified

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

Summary

This paper details the design, implementation, and public trial of autonomous golf cars deployed in Singapore’s Chinese and Japanese Gardens to promote public acceptance of autonomous vehicles. The research addresses the challenges of Mobility-on-Demand (MoD) systems, specifically focusing on vehicle rebalancing and the need for robust, safe autonomous shuttles that can operate in shared pedestrian environments. The primary motivation was to demonstrate technology maturity and gain user concurrence through a two-week public trial where visitors could book rides between ten stations. The system utilized two retrofitted Yamaha YDREX3 electric golf cars equipped with drive-by-wire controls for steering, braking, throttle, and gear selection. The hardware included a redundant power system with an auxiliary battery to prevent shutdowns during high-load operations, and multiple SICK LIDAR sensors for localization and obstacle detection. Computing was handled by two desktop-class PCs running Ubuntu and the Robot Operating System (ROS), protected by industrial racks with liquid cooling to withstand Singapore’s high heat and humidity. The software architecture comprised four modules: perception, planning, control, and external communication. Localization employed Adaptive Monte-Carlo Localization using synthetic LIDAR features, while moving object detection utilized Support Vector Machines to identify pedestrians. A key safety feature was the Dynamic Virtual Bumper (DVB), which adjusted the vehicle’s advisory speed based on the proximity of static and moving obstacles. Mission planning optimized routes to minimize travel distance and passenger wait times, while Vehicle-to-Vehicle (V2V) communication managed traffic on a single-lane bridge connecting the gardens. The trial demonstrated that the vehicles operated robustly over prolonged durations with small localization variance. The system successfully handled stochastic customer demand and complex environmental factors, including variable lighting and rain, due to LIDAR-based sensing. User surveys indicated high receptiveness to the technology. The intuitive interface, which allowed bookings via touchscreen or web, and safety features such as the removal of the steering wheel to prevent conflict with the autonomous motor, contributed to a positive user experience. The vehicles maintained safety through geofencing and emergency stop mechanisms accessible to both passengers and pedestrians. The significance of this work lies in its demonstration of a fully functional, safe, and user-friendly autonomous MoD system in a real-world public setting. It provides evidence that autonomous shuttles can operate reliably in pedestrian-heavy environments without dedicated infrastructure, addressing key barriers to adoption such as safety concerns and public acceptance. The study highlights the importance of robust hardware design for environmental resilience and intuitive human-machine interfaces for user trust, offering a blueprint for future autonomous public transit trials.

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.

• driverless ads