||Edge computing has emerged recently as a promis- ing supplement to cloud computing and is especially useful for resource-limited tactical networks. It enables devices to offload computationally intensive analytics tasks to nearby edge nodes. In this paper, we consider an edge cloud with multiple edge nodes and multiple edge users with analytics tasks arriving over time. Our goal is to maximise the overall social welfare (i.e., the difference between the value and the operational costs of all tasks). To make the system more robust, we propose to use decentralised reverse auctions for allocating tasks to resources in the edge cloud. By enabling competition between resource providers, these auctions ensure that the most suitable provider is chosen for a given task, but without the computational and communication overheads of a centralised solution. Moreover, such auctions deal naturally with highly dynamic systems, where nodes may appear or disappear over time. In order to derive effective bidding strategies for nodes, we propose a novel reinforcement learning algorithm that takes into account the status of a node and task characteristics, and that aims to maximise the node’s long-term revenue.