Using Node-RED for IIoT

With the release of an interface board for Raspberry Pi to control the industrial models built with fischertechnik. The review of the board is available at Didacta Advance Pi-F5 interface board.

Sure, the Raspberry Pi runs on Linux and thus could not offer real-time deterministic monitoring or be considered as a PLC.

However, Node-RED offers a nice graphic environment to develop the firmware that controls the industrial model.

The initial flow

The initial flow wasn't very pretty nor easy to read, so I opted for implementing a solution based on finite-state machine.

This method relies on an abstract model that lists a finite number of states —hence the name. Inputs can change the machine from one state to another: such changes are called transitions. A last element is the initial state the machine starts with.

I tried two nodes, node-red-contrib-finite-statemachine and node-red-contrib-fsm, later replaced by node-red-contrib-persistent-fsm.
.

Finite-state machine to the rescue

The model includes 5 states, an initial state, and 9 transitions of 6 kinds.

So the first step was to identify the states and the transitions, and draw the graph accordingly.

The node-red-contrib-finite-statemachine node requires to enter the states and transitions as a JSON sequence, but displays them with a nice graph.

So I opted for the node-red-contrib-fsm node where states and transitions are defined in a table.

The node-red-contrib-persistent-fsm node provides a similar table for input of the states and the transitions, displays them with a nice graph.

The flow consists on three parts:

The outputs part manages the motor, the blinking LED and the dashboard.

The FSM part defines the actions for each state.

Actions include setting the motor, turning the blinking LED on or off, and displaying a message.

The inputs part converts input into transitions for the FSM node.

Each event raised by a button, a phototransistor or a timer first goes through validation and then defines a transition sent to the FSM node.

Conclusion

Turning the initial flow into a finite-state machine was quick and easy. The major change is the split between the actions linked to a state and the inputs that trigger a transition.

This make the flow easier to read and maintain.