Application of extended Dempster-Shafer theory of evidence in accident probability estimation for dangerous goods transportation

Transportation of dangerous goods (DGs) is generally associated with significant levels of risk. In the context of DG transportation, risk refers to the likelihood of incurring the undesirable consequences of a possible accident. Since the probability of an accident in a link of a route might depend on a variety of factors, it is necessary to find a way to combine the pieces of evidence/probabilities to estimate the composite probability for the link. Instead of using the Bayesian approach, commonly used in the literature, which requires decision-makers to estimate prior and conditional probabilities and cannot differentiate uncertainty from ignorance, this paper presents a novel approach based on the extended Dempster-Shafer theory of evidence by constructing an adaptive robust combination rule to estimate the accident probability under conflicting evidence. A case study is carried out for the transportation of liquefied petroleum gas in the road network of Hong Kong. Experimental results demonstrate the efficacy of the proposed approach.