Cyber-Physical Systems (CPS) integrate physical processes, sensors, and embedded computers to facilitate advanced control systems such as autonomous cars and smart cities. Communication in CPS has tight constraints regarding reliability and latency, while traditional networking primitives can not guarantee these constraints. We base our thesis on the Chaos networking primitive which is a new paradigm that overcomes these limitations of traditional networking primitives. The current radio standard on which Chaos relies, is used more in science and industry, but less in everyday devices. We want to bring Chaos from the lab to the real world by porting the primitive to Bluetooth Low Energy (BLE). This thesis presents Chaos BLE: an implementation of the Chaos communication primitive on Bluetooth Low Energy. We characterised the capture effect on BLE and achieved accurate time synchronisation among nodes (< 2.5?s), both of which are key for the operation of the Chaos primitive. We validated our design and implementation on a 25 node BLE testbed that we have build at Delft University of Technology. In order to improve the performance of Chaos BLE and to mitigate channel interference, we propose a multichannel approach. The Chaos Multichannel primitive enables the network to use multiple channels in parallel, such that smaller sub-networks arise. Chaos Multichannel outperforms the single channel primitive in terms of reliability and latency. It achieves reliabilities close to 100% while finding a consensus among the nodes up to 2 times faster, compared to the single channel approach.