A patrol robot cycles through the same six checkpoints forever. On a fixed route it always visits them in the same order with the same dwell time at each stop, so the interval between two consecutive visits to any given checkpoint quickly converges to a constant. Once an observer has watched one or two full rounds, that interval is fully known — the "safe gap" right after the robot leaves is guaranteed and repeatable.
predicted_next = last_departure + median(observed_gaps)
exploit_time = last_departure + median(observed_gaps)
On a randomized patrol the checkpoint order is reshuffled every round and the dwell time at each stop is drawn from a random range, so the gap between consecutive visits to the same checkpoint varies round to round — occasionally the reshuffle even revisits the same checkpoint twice in quick succession. The exact same exploit_time formula, fed the same learned typical gap, now lands inside the robot's actual visit window unpredictably often — because no single learned number tracks the true next gap.
- Predicted next visit — countdown computed purely from history, exactly as an observer would estimate it. Watch it stay locked-on for the fixed route and drift for the randomized one.
- Attempt intrusion now — resolves instantly against the robot's real position: caught if it is within range of the targeted checkpoint at that moment.
- Auto-intruder — automatically fires one exploit attempt per round at the targeted checkpoint's just-learned safe gap, so the success-rate tally accumulates over many rounds without manual clicking.
Real-world relevance: unpredictable scheduling is a standard countermeasure in physical security — the same reason airport security and guard patrols intentionally vary timing, because a perfectly repeatable routine is a rehearsed vulnerability.