Multi-robots are often better choices to complete search-and-rescue tasks in some large-scale disasters,such as earthquakes,mine accidents,forest fires,etc.However,effects such as shadowing and fading for GNSS signals limit the positioning ability which is most important for search-and-rescue robots.To improve positioning availability and reliability,the article proposes a GNSS Peer-to-Peer cooperation positioning system for multi-robots search-and-rescue.The peers share GNSS-only data among neighbors as aiding information under light block scenario,and share both GNSS and terrestrial ranging data under deep indoor scenario.A particle filtering algorithm,the Monte Carlo numerical approximation of Bayesian filtering,is proposed to estimate position of peers utilizing both the prior information coming from robot motion model and posterior information provided by pseudo-range and terrestrial range observations,and the algorithm flow chart is presented.As a result,the acquisition time could be reduced and sensitivity could be improved for peers under light block scenario,and position could be solved under deep indoor scenario with fewer than 4 visible satellites.Simulation results show that the positioning error of particle filtering is less than that of Non-Bayesian filtering,and the error is about 5 meters for low-cost receivers.