@Michal-Mormon
S_SCENE_CONTROLLER can send messages V_SCENE_ON and V_SCENE_OFF. If you send those messages for the first time, domoticz will create new switch with the name 'scene'.

@ErrK I had similar problem with my hc-sr501 pir. It seems that some pirs are very sensitive to any electromagnetic noise. For me pir was reporting false positives each time nrf was sending data. You can easily test it: just remove any gw.send commands and solder led to pir output (of course through 1k resistor). When pir is detecting a motion output goes high so led will go on for the time set by pir potentiometer. In my case each sending (when for example battery level was reported each 30 min) resulted in false pir trigger.
It seems that this depends on specific device. I switched pirs between my sensors and this false triggering went to other node
Finally I solved this problem in software. I can share my sketch that is working well for two of my nodes.

Btw: there are many discussions on the web about those pir false triggering concerning remote radio nodes. I took the idea of software solution from there.

Of course in your case reasons can be different...

@Michal-Mormon
Yes, exactly! It looks like without scene_off sent from scene controller in domoticz it will be one-time action (since if this scene switch is switched on you cannot switch it on again). You can try to setup automatic switching it off in domoticz.

@Sweebee Depending on value of these pull-up resistors. I have read that they have about 50k, so with vcc=3v you will get 60u.

@Sweebee I see in the sketch, that you are enabling internal pull-up on PIR input. This means, that if PIR is not detecting movement and its output is set to zero, this pull-up resistor consumes 60uA (in the best case).

@rvendrame But I work only with pirs with ldo removed (and diode).

@ahmedadelhosni All those PIRs are built using the same chip. All all have 3,3v regulator onboard, so they always work with 3,3v. I don't understand why powering directly with 3,2v from battery makes so trouble.
Maybe this LDO regulator adds some additional stabilisation/filtering on power line, when powered with greater voltage...

@rvendrame I'm powering all with 2xAA baterries. On PIR I removed regulator. The problem with false trigerring is independent from voltage level. It is the same if I put old baterries (2,8V) or brand new (3,2V).
@fifipil909 I also suspect that it is somehow connected to powering noises and mcu sleeping. In my one case PIR is triggering each minute - and sleep time is also one minute.

@ErrK In my case PIR switches into HIGH for some seconds (depending on potentiometer), so reading after 100ms will give the same value.

@ErrK Unfortunately my solution is not 100% reliable. For 2 of my nodes it works well, but for 3rd node (the same hardware, the same sketch, only PCB is a little bi different - previous version, but difference only in dimension) - pir is false triggerring almost each minute (sometimes with 2 minutes delay with false triggering). And I checked - it is not caused by NRF sending. I have no idea whats going on.

@Michal-Mormon
Yes, exactly! It looks like without scene_off sent from scene controller in domoticz it will be one-time action (since if this scene switch is switched on you cannot switch it on again). You can try to setup automatic switching it off in domoticz.

@Michal-Mormon
S_SCENE_CONTROLLER can send messages V_SCENE_ON and V_SCENE_OFF. If you send those messages for the first time, domoticz will create new switch with the name 'scene'.

@ErrK I had similar problem with my hc-sr501 pir. It seems that some pirs are very sensitive to any electromagnetic noise. For me pir was reporting false positives each time nrf was sending data. You can easily test it: just remove any gw.send commands and solder led to pir output (of course through 1k resistor). When pir is detecting a motion output goes high so led will go on for the time set by pir potentiometer. In my case each sending (when for example battery level was reported each 30 min) resulted in false pir trigger.
It seems that this depends on specific device. I switched pirs between my sensors and this false triggering went to other node
Finally I solved this problem in software. I can share my sketch that is working well for two of my nodes.

Btw: there are many discussions on the web about those pir false triggering concerning remote radio nodes. I took the idea of software solution from there.

Of course in your case reasons can be different...

Some comments:

• You should include resistors (e.g. 100 ohm) between mc pin and transistor gate. Just to protect controller pin agains high current (when gate is charged).
• PWM output is available only for some pins. "On most Arduino boards (those with the ATmega168 or ATmega328), this function works on pins 3, 5, 6, 9, 10, and 11."
  So I don't think you can use pin 4 for green.
• Additionally - PWM on pins 5 and 6 is a little broken; arduino core libraries use the same timer for millis() and delay() functions. Refer to documentation:
  "This will be noticed mostly on low duty-cycle settings (e.g 0 - 10) and may result in a value of 0 not fully turning off the output on pins 5 and 6."
• R2 is not necessary. Just enable internal pull-up in controller. The same for R1.
• Your schematic is a little bit ugly There are missing may dots on signal connections.
  The rest is quite fine. I also plan to build similar node.