Researchers have proposed a new variant of the orienteering problem (OP) called OP-UTVR. This problem involves uncertain and time-varying rewards in real-world applications such as service robotics. Unlike traditional OP formulations that assume known rewards, OP-UTVR allows agents to estimate reward dynamics from observations and forecast future rewards. The authors have developed three planners with different planning horizons and online adaptivity, and provided theoretical bounds on their performance. They also introduced a benchmark for mobile service robots navigating in indoor environments. --- Why it matters: This work is relevant to researchers and engineers working on AI-powered service robotics, as it addresses the challenge of making informed routing decisions despite uncertain and changing rewards. Source: https://arxiv.org/abs/2608.18672