Wireless sensor-actuator networks (WSANs) are designed to collect and disseminate information using a physically distributed collection of wireless nodes and multi-hop protocols. WSANs are gaining rapid adoption in industrial automation and manufacturing applications due to their low deployment cost, robustness, and configuration flexibility. While the early success of industrial WSANs has been focused on monitoring applications, there are significant advantages, and also challenges, when WSANs are used in feedback control applications. Deployments of WSANs in control applications require careful design and testing of their network configurations and of the control algorithms employed, since undesired behaviors can arise from differences between expected and observed latency and reliability. This project creates a holistic approach to design and operate industrial wireless process control systems based on new closed-loop interactions between the controller of the industrial process and the WSAN manager. In this new closed loop, the controller, using a prediction model of the industrial process, estimates and compares the control performance loss associated to different network configurations. Accordingly, the WSAN manager estimates the quality of its internal links and adapts the network configuration to optimize the controller performance. At the core of this proposal is a holistic controller, which besides choosing the input signal for the physical process, selects a network configuration from a finite set of options with the goal of providing safety and performance guarantees. The holistic controller uses estimations of network status and physical process states to decide an appropriate network configuration, keeping in mind the inherent cost and delay associated to changes in scheduling and routing. The WSAN manager will, in turn, implement a new interface to communicate with the holistic controller, new mechanisms for efficient network reconfiguration, and new observers to estimate the probability of information delivery as a function of the different configurations and environmental conditions. The proposed feedback configuration between holistic controller and network manager enables industrial process control applications that are resilient against disturbances to both the physical plant and the wireless network. Furthermore, implementation in large-scale networks is enabled using a proposed bidirectional middleware, designed to transfer information between holistic controllers and network managers in real-time, exposing only the relevant features of each to the rest of the system. The results of this project will transform the way in which industrial wireless process control systems are designed, deployed, and operated, while establishing a new class of adaptive wireless control systems.