Logo
Home|Product|

Safety Guardrails

Safety Guardrails

Guardrails are critical for RL. It helps the agent gain autonomy through restrictions and allows it to write setpoints based on the most feasible outcome. Here’s how the guardrails help RL. 

Limits on agent actions
  • Action ranges for setpoints  
  • Limits on change in the action per step
Limits on process variables
  • Setting operational limits  
  • Optimizing resource usage
Human intervention
  • Alarms to operators   
  • On-off button (cascade mode)

By implementing guardrails, the RL system can enhance safety. Here’s how: 

  • Constantly monitoring

    The RL agent can alert early in case of any unexpected outcomes.

  • Constantly adapting

    The RL agent adjusts with changes in plant conditions, sensors, etc. to improve reliability by actively keeping the process in a desired state.

  • Better system understanding

    Improved process understanding by revisiting existing control, adding new alarms, and adding new setpoints tested by the RL agent.