Lifelong exploration of dynamic environments

We propose a novel spatio-temporal mobile-robot exploration method for dynamic, human-populated environments. In contrast to other exploration methods that model the environment as being static, our spatio-temporal exploration method creates and maintains a world model that not only represents the environment's structure, but also its dynamics over time. Consideration of the world dynamics adds an extra, temporal dimension to the explored space and makes the exploration task a never-ending data-gathering process to keep the robot's environment model up-to-date. Thus, the crucial question is not only where, but also when to observe the explored environment. We address the problem by application of information-theoretic exploration to world representations that model the environment states' uncertainties as probabilistic functions of time. The predictive ability of the spatio-temporal model allows the exploration method to decide not only where, but also when to make environment observations. To verify the proposed approach, an evaluation of several exploration strategies and spatio-temporal models was carried out using real-world data gathered over several months. The evaluation indicates that through understanding of the environment dynamics, the proposed spatio-temporal exploration method could predict which locations were going to change at a specific time and use this knowledge to guide the robot. Such an ability is crucial for long-term deployment of mobile robots in human-populated spaces that change over time.