Cooperative navigation (CN) is a widespread technique to have efficient navigation of intelligent vehicles. Nonetheless, the CN strategies need to be more consistent in estimating and managing in-road risks. This paper outlines a flexible CN scheme for multiple unmanned ground vehicles (MUGVs) system to deal with such critical cooperative system. With its relative low execution time, the probability collectives (PC) algorithm has succeeded at generating fast and feasible solutions to cross intersections and roundabouts (Philippe et al. 1928–1934, 2019). However, the PC is still sensitive to uncertainty in the navigation process, which highlights the need to adopt several safety margins. This work focuses on balancing between the high-quality cooperative optimization and acceptable computational speed. Thus, a reliable risk management strategy is proposed by introducing a novel ε-constraint PC method. A real-time communication mechanism is suggested for a distributed system to avoid invalid behavior due to inconsistency. The novel ε-PC based navigation strategy allows the vehicles to adapt their dynamics and react to unexpected events while respecting real-time constraints. One finding appears to be well substantiated by the typical common-yet-difficult scenarios in intensive simulations. The -PC method can ensure collision-free behaviors and reserve at least 1.5s of reaction time for vehicles’ safety insurance.