Identifiability of car-following dynamics

The advancement of in-vehicle sensors provides abundant datasets to estimate parameters car-following models that describe driver behaviors. question parameter identifiability such (i.e., whether it is possible infer its unknown from the experimental data) a central system analysis question, and yet still remains open. This article presents both structural practical on four common models: i) constant-time headway relative-velocity (CTH-RV) model, ii) optimal velocity model (OV), iii) follow-the-leader (FTL) iv) intelligent (IDM). carried out using differential geometry approach, which confirms that, in theory, all tested systems are structurally locally identifiable, i.e., can be uniquely inferred under almost initial condition admissible inputs by observing space gap alone. In setting, we propose an optimization-based numerical direct test determine given specific setup (the conditions input known). conclusively finds distinct CTH-RV FTL not identifiable trajectory.