In indoor or urban applications, a moving robot with wireless communications will experience multipath fading. This causes rapid signal strength variations due to interfering reflections of the radio signal. By making short stops at positions with high signal-to-noise ratio (SNR), the robot can trade trajectory tracking accuracy for increased link quality. This represents a type of opportunistic communication-aware motion planning. We propose two novel strategies for improving the link capacity or throughput when either the robot has full knowledge of how the SNR varies along the trajectory, or when only the SNR distribution is known or can be estimated. In the latter case, this leads to an optimal stopping problem over a finite horizon. Both cases are analyzed for independent as well as correlated SNR samples, and a bounded maximum trajectory tracking error. We derive the resulting SNR distributions for the proposed strategies and use them to show how the expected capacity and throughput vary with the allowed tracking error. The results are confirmed by simulations and experiments. Experiments in six different locations validate the communication model and show that the proposed motion planning is robust to non-static fading and can yield throughput improvements of more than 100%.