Nowadays, wireless sensor networks (WSNs) have become omnipresent. Also location-based services become more and more popular. In this context, energy- and location-awareness are essential properties of modern sensor networks. Addressing low-power positioning, received signal strength indicator (RSSI)-based direction finding is a prospective approach for WSNs providing location-based services. However, like with all radio-based localization techniques, RSSI-based direction finding is prone to multipath propagation. In this paper, a probabilistic approach to multipath mitigation in RSSI-based direction estimation is presented. The presented approach is verified by Monte Carlo simulations. Simulations reveal the capability of zero-mean estimates for arrival angle in presence of multipath. Furthermore, the impact on the position estimation with recursive Bayesian filters is assessed. Trajectory estimation in a multipath setting shows promising results. Errors arising from multipath propagation can be minimized to a large extent.