If and When a Driver or Passenger is Returning to Vehicle: Framework to Infer Intent and Arrival Time

TL;DR

This paper introduces a probabilistic Bayesian framework to predict if and when a driver or passenger will return to their vehicle using partial location data, effectively capturing long-term intent dependencies with low complexity.

Contribution

It presents a novel Bayesian approach for intent and arrival time prediction that requires minimal training and models long-term trajectory dependencies.

Findings

Effective in predicting return intent and timing from partial data

Low training requirements and computational efficiency

Demonstrated success with two example scenarios

Abstract

This paper proposes a probabilistic framework for the sequential estimation of the likelihood of a driver or passenger(s) returning to the vehicle and time of arrival, from the available partial track of the user location. The latter can be provided by a smartphone navigational service and/or other dedicated (e.g. RF based) user-to-vehicle positioning solution. The introduced novel approach treats the tackled problem as an intent prediction task within a Bayesian formulation, leading to an efficient implementation of the inference routine with notably low training requirements. It effectively captures the long term dependencies in the trajectory followed by the driver/passenger to the vehicle, as dictated by intent, via a bridging distribution. Two examples are shown to demonstrate the efficacy of this flexible low-complexity technique.