From a Competition for Self-Driving Miniature Cars to a Standardized Experimental Platform: Concept, Models, Architecture, and Evaluation
This work provides a standardized platform for researchers in autonomous vehicles to conduct riskless and resource-efficient experiments, though it is incremental as it builds on existing simulation and competition insights.
The authors tackled the challenge of bridging simulation-based algorithm development and real-world autonomous vehicle testing by adapting a simulation platform for self-driving miniature cars and designing a standardized experimental platform, resulting in a reusable, low-cost hardware architecture that enables efficient experiments without needing large test tracks.
Context: Competitions for self-driving cars facilitated the development and research in the domain of autonomous vehicles towards potential solutions for the future mobility. Objective: Miniature vehicles can bridge the gap between simulation-based evaluations of algorithms relying on simplified models, and those time-consuming vehicle tests on real-scale proving grounds. Method: This article combines findings from a systematic literature review, an in-depth analysis of results and technical concepts from contestants in a competition for self-driving miniature cars, and experiences of participating in the 2013 competition for self-driving cars. Results: A simulation-based development platform for real-scale vehicles has been adapted to support the development of a self-driving miniature car. Furthermore, a standardized platform was designed and realized to enable research and experiments in the context of future mobility solutions. Conclusion: A clear separation between algorithm conceptualization and validation in a model-based simulation environment enabled efficient and riskless experiments and validation. The design of a reusable, low-cost, and energy-efficient hardware architecture utilizing a standardized software/hardware interface enables experiments, which would otherwise require resources like a large real-scale test track.