In an environment surrounded by obstacles, such as a mountain valley, legacy GPS receiver architectures often experience degraded accuracy due to dense multipath and/or signal masking. To mitigate challenging effects, we propose Multi- Receiver Direct Position Estimation (MR-DPE). MR-DPE is based on existing works in Direct Position Estimation (DPE), which estimates the navigation solution in the position domain directly. This approach provides deep inter-channel couplings and information redundancy. In MR-DPE, multiple DPE receivers are organized into a receiver network with fixed antenna baselines. A unified positionvelocity- time (PVT) solution is then synthesized by aggregating the measurements taken at the individual receivers. The robustness of MR-DPE is threefold: the measurement and the geometric redundancies originated from the multi-receiver framework, as well as the intrinsic robustness inherited from the individual DPE receivers. We implemented MR-DPE on our PyGNSS software-defined radio platform. The MR-DPE architecture was validated through a series of full-scale flight tests. We demonstrated the improved robustness of MR-DPE against multipath and terrain masking.