This paper addresses problems of large planning time and cost uncertainty for informative path planning of a mobile sensor where the location of sensor deployment is different of that of an operational area. The first problem is that the cost has no term dependent on sensor state before arriving at the operational area and it causes large planning time. The information of the state of interest dissipates over time during the planning time and it degrades performance of sensing operation. The other problem is that the cost is dependent on the parameters to be estimated. To assess the cost, the target state in the future should be predicted by integrating the system model based on noisy initial estimate. The limitation of the informative path planning has a greater impact on performance in this specific problem. A strategy to cope with these problems is to devise a real-time path planning algorithm by using online optimization. The proposed algorithm is divided into two phases; determining the path to the boundary of the operational area and guiding the sensor by an informative potential field in the area. Detailed analysis on performance of the proposed algorithm compared to an optimal solution by nonlinear programming is given. The simulation results have demonstrated that the proposed algorithm can cope with performance degradation observed in the optimal solution.