Sensor control system for controlling a sensor network
A sensor control system ( 202 ) for managing at least a first set of one or more sensors ( 101 ) for monitoring a first domain of an industrial process and a second set of one or more sensors ( 102 ) for monitoring a second domain of the industrial process, wherein the sensor control system ( 202 ) comprises at least a first reinforcement learning, RL, agent (A 1 ) and a second RL agent (A 2 ), wherein the first and second RL agents were trained using reinforcement learning and a process graph ( 196 ) representing the industrial process.
1. A method performed by a sensor control system for managing at least a first set of one or more sensors for monitoring a first domain of an industrial process and a second set of one or more sensors for monitoring a second domain of the industrial process, wherein the sensor control system comprises at least a first reinforcement learning (RL) agent and a second RL agent, wherein the first and second RL agents were trained using reinforcement learning and a process graph representing the industrial process, the method comprising:
the sensor control system receiving sensor data from the first set of one or more sensors;
the sensor control system using the received sensor data and the process graph to decide whether or not to reconfigure the first set of sensors and/or the second set of sensors; and
the sensor control system providing configuration information to the first set of sensors and/or the second set of sensors as a result of the sensor control system deciding to reconfigure the first set of sensors and/or the second set of sensors,
wherein the first and second domains are defined based on sensor locations; or
wherein the first and second domains are defined based on functional similarities.
2. The method of claim 1 , wherein
the first set of sensors are configured to monitor a first workstation, and
the second set of sensors are configured to monitor a second workstation.
3. The method of claim 2 , wherein receiving sensor data from the first set of one or more sensors comprises the first RL agent receiving the sensor data from the first set of sensors.
4. The method of claim 3 , wherein using the received sensor data and the process graph to decide whether or not to reconfigure the first set of sensors and/or the second set of sensors comprises:
the first RL agent detecting an anomaly with respect to the first workstation based on the received sensor data; and
the first RL agent, as a result of detecting the anomaly with respect to the first workstation, obtaining information about a third workstation that is monitored by a third set of sensors.
5. The method of claim 4 , further comprising the first RL agent using the obtained information about the third workstation to decide whether or not to reconfigure the second set of sensors.
6. The method of claim 1 , further comprising:
training the first RL agent using: i) the process graph, ii) sensor data, and iii) communication capacity information; and
training the second RL agent using: i) the process graph, ii) the sensor data, and iii) the communication capacity information.
7. The method of claim 6 , wherein training the first and second RL agents comprises:
performing a first training phase where the first RL agent is trained to optimize a first local optimization function and the second RL agent is trained to optimize a second local optimization function; and
performing a second training phase where the first and second RL agents are trained to optimize a predefined weighted sum of local objective functions.
8. The method of claim 1 , wherein the first and second domains are defined based on sensor locations.
9. The method of claim 1 , wherein the first and second domains are defined based on functional similarities.
10. The method of claim 1 , wherein providing the configuration information to the first set of sensors and/or the second set of sensors comprises transmitting the configuration information to a sensor gateway that is configured to relay the configuration information to the first set of sensors and/or the second set of sensors.
11. A non-transitory computer readable storage medium storing a computer program comprising instructions which when executed by processing circuitry of a sensor control system causes the sensor control system to perform the method of claim 1 .
12. A sensor control system for managing at least a first set of one or more sensors for monitoring a first domain of an industrial process and a second set of one or more sensors for monitoring a second domain of the industrial process, the sensor control system comprising:
a first reinforcement learning, (RL) agent; and
a second RL agent, wherein
the first and second RL agents were trained using reinforcement learning and a process graph representing the industrial process, and
the sensor control system is operable to:
i) receive sensor data from the first set of one or more sensors;
ii) use the received sensor data and the process graph to decide whether or not to reconfigure the first set of sensors and/or the second set of sensors; and
iii) provide configuration information to the first set of sensors and/or the second set of sensors as a result of deciding to reconfigure the first set of sensors and/or the second set of sensors.
13. A sensor control system for managing at least a first set of one or more sensors for monitoring a first domain of an industrial process and a second set of one or more sensors for monitoring a second domain of the industrial process, the sensor control system comprising:
a receiver for receiving sensor data from a first set of one or more sensors;
processing circuitry; and
a memory, the memory containing instructions executable by the processing circuitry, wherein the sensor control system is configured to:
use the received sensor data and a process graph representing the industrial process to decide whether or not to reconfigure the first set of sensors and/or a second set of sensors; and
provide configuration information to the first set of sensors and/or the second set of sensors as a result of the sensor control system deciding to reconfigure the first set of sensors and/or the second set of sensors,
wherein the first and second domains are defined based on sensor locations; or
wherein the first and second domains are defined based on functional similarities.
14. The sensor control system claim 13 , further comprising:
training a first reinforcement learning (RL) agent using: i) the process graph, ii) sensor data, and iii) communication capacity information; and
training a second RL agent using: i) the process graph, ii) the sensor data, and iii) the communication capacity information.
15. The sensor control system claim 14 , wherein training the first and second RL agents comprises:
performing a first training phase where the first RL agent is trained to optimize a first local optimization function and the second RL agent is trained to optimize a second local optimization function; and
performing a second training phase where the first and second RL agents are trained to optimize a predefined weighted sum of local objective functions.
16. The sensor control system claim 14 , wherein
the first set of sensors are configured to monitor a first workstation,
the second set of sensors are configured to monitor a second workstation,
receiving sensor data from the first set of one or more sensors comprises the first RL agent receiving the sensor data from the first set of sensors,
using the received sensor data and the process graph to decide whether or not to reconfigure the first set of sensors and/or the second set of sensors comprises:
the first RL agent detecting an anomaly with respect to the first workstation based on the received sensor data; and
the first RL agent, as a result of detecting the anomaly with respect to the first workstation, obtaining information about a third workstation that is monitored by a third set of sensors, and
the first RL agent is configured to use the obtained information about the third workstation to decide whether or not to reconfigure the second set of sensors.
17. The sensor control system claim 13 , wherein the first and second domains are defined based on sensor locations.
18. The sensor control system claim 13 , wherein the first and second domains are defined based on functional similarities.
19. The sensor control system claim 13 , wherein providing the configuration information to the first set of sensors and/or the second set of sensors comprises transmitting the configuration information to a sensor gateway that is configured to relay the configuration information to the first set of sensors and/or the second set of sensors.