Optimizing Task Management in Socially Assistive Robots Using an MDP Framework

Socially assistive robots (SARs) must decide when and how to encourage task completion while maintaining sustained engagement. Beyond assistive and homecare contexts, similar decision-making challenges arise in smart mobility and logistics ecosystems, where autonomous agents must prioritize tasks, coordinate with human operators, and adapt to dynamic operational conditions. This research presents a Markov Decision Process (MDP) framework for task (drive) management that enables a robot to optimize its decision-making strategy by considering both task priorities and human engagement levels. The framework comprises state variables representing drives and human interest, transition probabilities derived from a Bayesian network, and a reward function that balances urgency with sustained interaction quality. A case study demonstrates how optimal policies can be obtained using value iteration. While illustrated in a socially assistive setting, the proposed framework is equally applicable to warehouse robotics, airport service robots, and mobility service platforms that require adaptive task allocation under uncertainty. The results highlight the potential of MDP-based task management to enhance long-term operational efficiency and adaptive human–robot coordination in smart mobility and logistics environments.