Heat-Diffusion (HD) routing is a recently-developed queue-aware routing policy for multi-hop wireless networks inspired by Thermodynamics. In prior theoretical studies, it has been shown that HD routing guarantees throughput optimality, minimizes a quadratic routing cost, minimizes queue congestion on the network, and provides a trade-off between routing cost and queueing delay that is Pareto-Optimal. While striking, these guarantees are based on idealized assumptions (including global synchronization, centralized control, and infinite buffers) and heretofore have only been evaluated through simplified numerical simulations. We present here the first practical decentralized Heat-Diffusion Collection Protocol (HDCP) for wireless sensor networks and detail its implementation on Contiki OS. We present a thorough evaluation of HDCP based on real testbed experiments, including a comparative analysis of its performance with respect to the state of the art Collection Tree Protocol (CTP) and Backpressure Collection Protocol (BCP) for wireless sensor networks. We find that HDCP has a significantly higher throughput region and greater resilience to interference compared to CTP. However, we also find that best performance of HDCP is comparable to the best performance of BCP, due to the similarity in their neighbor rankings, which we verify through a Kendall’s-Tau test.