Off-Road Autonomy Validation Using Scalable Digital Twin Simulations Within High-Performance Computing Clusters

TL;DR

This paper proposes leveraging scalable digital twin simulations within high-performance computing (HPC) clusters to provide a scalable and efficient means to validate off-road autonomy algorithms, enabling rapid iteration and testing of autonomy algorithms under various conditions.

Abstract

Off-road autonomy validation presents unique challenges due to the unpredictable and dynamic nature of off-road environments. Traditional methods focusing on sequentially sweeping across the parameter space for variability analysis struggle to comprehensively assess the performance and safety of off-road autonomous systems within the imposed time constraints. This paper proposes leveraging scalable digital twin simulations within high-performance computing (HPC) clusters to address this challenge. By harnessing the computational power of HPC clusters, our approach aims to provide a scalable and efficient means to validate off-road autonomy algorithms, enabling rapid iteration and testing of autonomy algorithms under various conditions. We demonstrate the effectiveness of our framework through performance evaluations of the HPC cluster in terms of simulation parallelization and present the systematic variability analysis of a candidate off-road autonomy algorithm to identify potential vulnerabilities in the autonomy stack's perception, planning and control modules.

Figures (7)

• Figure 1: Interactive web application visualizing the variability testing of autonomous RZR digital twin across different weather conditions and times of the day using parallel simulations in the cloud.
• Figure 2: Autonomy-oriented digital twin of the Polaris RZR PRO R 4 ULTIMATE: (a) physical RZR in the real world, (b) digital twin of RZR in AutoDRIVE Simulator, and (c) simplified representation of the vehicle dynamics and sensor simulation models.
• Figure 3: A feature-rich virtual environment for validating a wide spectrum of autonomy algorithms. The depicted environment is spread across 2$\times$2 km$^2$ area and comprises various driving segments such as a structured downtown, a mountain pass, a long stretch of highway, and over 4 km of dirt road. The virtual autonomous RZR was deployed in the dirt road zone.
• Figure 4: Overview of the HPC deployment framework for running AutoDRIVE Simulator in the cloud.
• Figure 5: Candidate off-road autonomy algorithm chosen for this study in action: vision-guided autonomous emergency braking for the RZR digital twin in off-road environment.