Case Studies

Case Study (June 10)

From Compact to Combat: Smart Hinges for Robust Structural Deployment

Led by Lorenzo Valdevit (PI, UCI) & Evgueni Filipov (Co-PI, UM)

Abstract TBA.

Case Study (June 11)

Adaptive Testing Without Losing the Plot: Ensuring Equivalency in Parameter-Driven Test Generation Led by Daniel Carruth (PI, MSU) & Christopher Hudson (Co-PI, MSU)

Current evaluation methods for autonomous ground vehicles (AGVs) often rely on highly specific test setups that improve repeatability across platforms, but limit insight into how systems respond to underlying driving challenges. To address this, we use the Mississippi State University Autonomous Vehicle Simulator (MAVS) to evaluate a Polaris MRZR, Subaru Forester, and Army truck in five test cases: Straight Line Jersey Barrier, Random Obstacle, Slalom, Vegetation Override, and Emergency Braking. To represent real operating conditions, we vary the visual appearance of critical scene elements and scale obstacle spacing and path dimensions relative to vehicle size. These modifications preserve the perceptual, geometric, and traversal challenges, enabling fair comparison across vehicles. Rather than relying on binary pass/fail outcomes, we execute large simulation ensembles with increasing environmental difficulty to identify system limits in perception, traversal, and control stability. Results are integrated into a GitLab CI/CD pipeline for automated analysis and reproducibility. This methodology enables scalable AGV evaluation using statistically grounded performance envelopes, while incorporating realistic scenario variation that maintains equivalence in underlying challenges, ensuring fair comparison and capturing behavior under representative operating conditions.