Probabilistic routing-based injury avoidance navigation framework for pedalcyclists
MetadataShow full item record
Easing traffic congestion in urban areas is a multifactorial challenge requiring continuous effort. Encouraging commuters to use alternative transportation such as bicycles is one simple way to reduce traffic congestion by reducing the number of motor vehicles in traffic. However, commuters involved in bicycle crashes sustain much more severe injuries than commuters involved in similar motor vehicle crashes. This risk of injury presents an obstacle to convincing commuters to use bicycles that improving roadway safety can help to overcome. Conventional roadway safety improvements are made at the local level based on recommendations from studies analyzing crash and traffic data. While changes made using this approach can increase safety in the long term, the cost-benefit analysis of implementing these changes limits their wide-spread implementation. This reactive approach has several drawbacks: Few road ways are improved, implementation is delayed, and it does not focus on safety benefits for bicycle commuters. Recent advances in data mining present a cost-effective proactive alternative that uses traffic and crash data readily available to the general public. In this paper, a navigation framework is proposed that allows analysis of this data to recommend safer routes to bicycle commuters. Machine learning techniques are applied to the injury severity metric presented in the crash data to develop a risk model for estimating route safety. A framework is created to obtain multiple routes and generate their safety scores by applying the risk model. An end-user navigation application was implemented to present multiple routes along with their safety scores. This proactive approach enables commuters to switch to bicycling by providing personalized safety recommendations without the limitations of the conventional approach.