The Internet of Things (IoT), which builds a large network of billions and trillions of ''things'' that interact with each other, faces many technological and application challenges. As many devices are interconnected with each other, there is a high possibility of congestion in the network. In critical IoT applications, on-time delivery of packets is important. It is therefore important to minimize network congestion, which can cause delay and packet drop when left unnoticed. Considering the above problem, a Reliable Path Routing (RPR) protocol for congestion control in critical IoT applications is proposed in this paper. To avoid congestion, the next-hop node to transfer the data should be thoughtfully made. For this purpose, the RPR protocol calculates Node Selecting Factor (NSF) based on the buffer occupancy level, interference, path survivability, and congestion level of the next-hop nodes. The node with the highest NSF value is selected to transfer the packets. Furthermore, weighting coefficients are given to each selection criteria out of which buffer occupancy level is given higher weightage than others so it reduces the delay suffered by packets. From the simulation results, it is evident that the RPR protocol decreases delay by an average of 38% and increases the packet delivery ratio by an average of 7% than SPR, SGEAR, CoAR, and CDTMRLB protocols.