← Back to the chapter · Previous detail
Technical detail · ACE and Shared Context Sensing
Freshness gates savings; correlation can skip a sensor
14.16 Freshness Gates The Savings, Correlation Skips The Sensor
Suppose 20 components request the same context each minute. A 60-second window permits one sense and 19 lookups: 0.34 + 19 × 0.016 = 0.644 mA-s/min. A workload yielding ten senses and ten lookups under a 2-second window costs 10 × 0.34 + 10 × 0.016 = 3.56 mA-s/min. That is still below 20 independent senses, but exceeds the longer-window cost by more than five times. The second mix requires appropriately spaced or clustered arrivals; the request count and TTL alone do not determine the hit rate.
Correlation supplies a separate lever. A learned rule might use an accelerometer’s “stationary” result as evidence that location has not changed, avoiding GPS. A GPS module tracking at roughly 25 mA consumes about 250 mA-s, or 0.069 mA-h, during a 10-second fix. A low-power accelerometer check at about 10 μA for 100 ms consumes roughly 0.001 mA-s. If the device is stationary 70% of the time and would otherwise take 100 fixes daily, avoiding 70 saves about 70 × 250 mA-s = 4.9 mA-h/day before proxy and validation overhead.
That rule remains conditional: “stationary implies same place” can drift and conceal real movement. Validate the relationship and keep a direct fallback. A short window collapses reuse; a long one risks stale action. Freshness is an energy-control parameter as well as a data-quality limit.