A stuck robot should proactively seek help rather than stop silently
Aliases: robot help-seeking · stuck detection · proactive assistance request
What it is
Proactive help-seeking by robots means explicitly requesting assistance when progress is impossible, recovery is exhausted, or execution is abnormally stalled. Silent cessation makes safe waiting, computation, completion, disconnection, and physical obstruction look identical, forcing continuous patrol and diagnosis.
Why it happens
Planning loops, absent displacement, actuator load, and sustained low confidence often expose blockage before a person notices. Connecting this evidence to a request turns passive monitoring into event-driven collaboration. Oversensitive thresholds cause interruptions; late thresholds extend downtime and permit repeated hazardous attempts.
Studying it
Trigger timing, failure type, and staff workload can be varied while measuring discovery delay, downtime, successful recovery, unnecessary interruption, and later alert response. Legitimate long computation must be separated from no progress, and attempted recoveries logged. Field logs establish real event frequency after safe fault injection.
Where it stops holding
Immediate hazards require safe stop without waiting for help, while brief self-recovering variation may not justify interruption. When nobody is available or communication fails, the robot still needs a safe terminal state. Help-seeking must not institutionalise a design defect as unpaid local labour.
Applying it
- Define stalled evidence for position, task progress, planning loops, and actuator load, allowing only bounded safe self-recovery.
- Stop consequence-amplifying action, show waiting-for-help state, and notify a capable role.
- Inject failures and measure detection, repeated attempts, arrival, and recovery; route recurrent requests into root-cause remediation.