Distribution-Free Probabilistic Environmental Control Barrier Functions with Application to Adaptive Cruise Control
Year: 2026
Type: preprint
Abstract: Safety control under environmental uncertainty is challenging when the uncertainty distribution is unknown and reliable support bounds are unavailable. This paper introduces CPP-ECBF, a distribution-free approach that combines conformal predictive programming with probabilistic environmental control barrier functions. For separable environmental safety functions and i.i.d. uncertainty, CPP-ECBF converts the probabilistic safety condition into a sample-based deterministic control constraint that provides finite-horizon safety with prescribed probability $1-ε$ and calibration confidence $1-\tildeβ$. For adaptive cruise control, an offline conformal uncertainty bound yields an explicit linear upper bound on the ego-vehicle acceleration, which is enforced by a quadratic-programming safety filter. In 10,000 Monte Carlo trials, 99.97% of trials satisfied the finite-horizon safety criterion, exceeding the prescribed confidence of 85%. Moreover, CPP-ECBF reduced the conservatism of worst-case robust enforcement, increasing the admissible acceleration upper limit by $5.90\,\mathrm{m/s^2}$ on average.
Source: arXiv (Cornell University)
Institution: Hanyang University
Language: English
Subfield: Computational Theory and Mathematics
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7235239049
