Rami Abousleiman and Romi Boimer
Bayesian Inference, Electric Vehicles, Optimization, Routing
Estimating the destination point or the intended travel destination is imperative for an improved passenger-vehicle experience. For example, this feature helps in improving power train efficiencies and electric range estimations. It can also help with focused content through targeted ads, or assist with road hazard alerts that enable preemptive situation mitigation. In this work, we present a Bayesian Inference based approach that learns from driving history. The model utilizes only few features such as time and current location to predict the destination point. To mitigate the passengers' behavioral changes, we add an evaporation rate factor so that older recordings have less influence on the model than newer ones. The evaporation rate also allows for a more evolved model that avoids getting stuck in local minimums. Results show that with less than 200 recordings we are able to successfully predict more than 85% of the travel destination points even with the incorporation of passengers' habitual changes
Important Links:
Go Back
