Energy & Power · Study deck

ACE and Shared Context Sensing

Picture three phone apps waking the same motion sensor for the same answer.

Battery Bruno is your guide for this deck.

contextsystem

Battery Bruno, the module guide, in a scene from this chapter.

After studying this chapter

Learning objectives

You will be able to:

Explain: Planning should change the sensor order, not merely rename an always-sense baseline. Figure: Normalised energy bars compare baseline compares that baseline with exhaustive and heuristic choices across four users.
Explain: This recurrence explains why “always choose the cheapest sensor first” can fail: a cheap sensor whose result rarely resolves the query merely adds its own cost before the expensive fallback.
Explain: The bars show that context names conceal different sensing costs: walking and driving can share a low-cost motion source, while meeting or working inference activates more expensive inputs.

Major section

Start With One Shared Fact

A broker means a service that receives information and passes it to the right users or systems.

This proves one shared context rule, not every privacy or accuracy case; the deeper sections cover the ACE roles, planning, inference, reuse, and limits.
If one sensor reading can answer several questions, repeating the same sensing work wastes energy.
Ownership, freshness, correlation, and validation decide whether shared context is an energy saving or a stale shortcut.

Major section

Shared Context Sensing

The cached value is not just a raw answer; it is an evidence record.

Several applications can reuse a result at near-zero lookup cost only when the cached record still satisfies their purpose.
A location tracker may tolerate an older estimate than a safety rule, so reuse is approved per consumer contract.

ACE architecture for shared context: multiple applications consume shared context values through ACE middleware and an inference cache rather than each sensing independently.

Major section

Cross-App Context Correlations

ACE becomes more powerful when it can infer one context attribute from another.

Confidence tells whether the rule is reliable when the antecedent is true.
A rare rule with high confidence may still be useful for a niche event; a common rule with weak confidence should not drive high-consequence decisions.

ACE rule mining gates showing observed context history, candidate rules, support filter, confidence filter, consequence threshold, and accepted rule set.

Major section

Cross-App Context Correlations (continued)

A rule observed twice in a month should not control an energy policy without more evidence.

Their relationship shows which part of the claim still needs evidence.
The next checkpoint: Candidates, adds a distinct review condition, while if-then rules adds a distinct review condition.
An old Driving = true value should not infer current location.

Major section

Cross-App Context Correlations (continued)

Together those labels show that no-panel ACE rule mining gates showing observed context history, candidate rules, support filter, confidence filter, consequence threshold, and accepted rule set.

Fallback gate: Every rule needs a direct sensing or conservative fallback when confidence drops, the context changes, or validation detects drift.
Planning should change the sensor order, not merely rename an always-sense baseline. Figure: Normalised energy bars compare baseline compares that baseline with exhaustive and heuristic choices across four users.
For Ada the exhaustive plan is 31 and the heuristic plan is 38; for Ben they are 42 and 47.

Normalised energy bars compare baseline, exhaustive, and heuristic ACE sensing plans for four users.

Major section

Sensing Planner

Value is fresh, trusted, and allowed for this consumer.

Stale context can silently make the wrong policy look cheap.
Rule confidence clears the threshold and the proxy is cheaper than target sensing.
The proxy may drift when user behavior or deployment patterns change.
A low-power sensor can update a rule antecedent before using an expensive target sensor.

Measured energy per sense in millijoules for ten context attributes, grouped by the sensor(s) each one requires.

Major section

Sensing Planner (continued)

Proxy sensing can become wasted work if the proxy is weak.

Cache and inference fail, or the decision needs direct evidence.
The bars show that context names conceal different sensing costs: walking and driving can share a low-cost motion source, while meeting or working inference activates more expensive inputs.
The planner should also record the rejected options.

ACE architecture for shared context sensing, from three application requests through cache, planner, contexters, rule history, and evidence outputs.

Major section

Sensing-Planner Complexity and a Practical Heuristic

A conditional sensing plan is a decision tree.

Each internal node either tests a cached attribute, wakes a sensor, or runs an inference; each outgoing edge represents a possible result.
The planner's job is to find a tree that minimizes expected energy without violating those correctness gates.
At a terminal leaf, $J=0$ only when the available evidence is sufficient to answer.

Major section

Sensing-Planner Complexity and a Practical Heuristic (continued)

Exact optimal decision-tree and related sensing-plan formulations contain known NP-hard problems.

This recurrence explains why “always choose the cheapest sensor first” can fail: a cheap sensor whose result rarely resolves the query merely adds its own cost before the expensive fallback.
Outcome probabilities must come from representative traces; invented confidence values only make the optimizer precisely wrong.
In production, re-plan when energy costs, cache-hit rates, or correlations drift.

ACE request fallback from fresh cache through association-rule inference to progressively more expensive proxy sensing.

Major section

Sensing-Planner Complexity and a Practical Heuristic (continued)

With $n$ optional attributes there are already $2^n$ possible subsets, $n!$ possible full orderings, and a branch can expose a different remaining choice after every outcome.

That statement applies to the general planning problem; a small, constrained instance can still be searched exactly.
The planner is an adaptive policy built from evidence, not a one-time static ordering of sensors.
Otherwise that branch must pay for another test or fall back to the authoritative sensor.

ACE routes application context requests through an inference cache, rule miner, sensing planner, and contexters, returning cached or inferred values when valid and sensing when needed.

Deck summary

Key takeaways

A broker means a service that receives information and passes it to the right users or systems.

The cached value is not just a raw answer; it is an evidence record.
ACE becomes more powerful when it can infer one context attribute from another.
A rule observed twice in a month should not control an energy policy without more evidence.
Together those labels show that no-panel ACE rule mining gates showing observed context history, candidate rules, support filter, confidence filter, consequence threshold, and accepted rule set.

Retrieval practice

Recall check 1 of 5

Battery Bruno says: answer from memory, then check your reasoning.

Q1App A sensed Walking=true 45 seconds ago with a TTL of 120 seconds and confidence 0.94. App B now requests Walking for a low-consequence activity summary with minimum confidence 0.80. What should an ACE-style broker do?

AAlways resense because a different app made the request

BReuse the cached value and record App B as a consumer

CInfer AtHome because App B's Walking request is overridden by the device's location evidence.

DExtend the TTL automatically to improve hit rate

Show answer

Answer: B ACE cache reuse is allowed when the context value is fresh, trusted, permitted, and adequate for the requesting decision.

Retrieval practice

Recall check 2 of 5

Battery Bruno says: answer from memory, then check your reasoning.

Q2How should the team treat this rule in the lighting control decision?

APromote it to reduce motion-sensor wakeups

BHold or reject the rule for lighting

CUse it only for high-consequence decisions

DIncrease cache duration until support improves

Show answer

Answer: B Rule promotion depends on support, confidence, freshness, and consequence.

Q3What learns supported relationships that can avoid a direct sensor read under thresholds?

ARule miner

BApp request

CInference cache

DSensing planner

Show answer

Answer: A It learns supported relationships that may replace a direct sensor read under explicit thresholds.

Retrieval practice

Recall check 3 of 5

Battery Bruno says: answer from memory, then check your reasoning.

Q4Four subsystems each need the same occupancy context once a minute. What makes a shared inference cache save energy compared with each subsystem sensing on its own?

AThe cache lowers the sensor's physical current during each reading.

BThe cache removes the need to sense again under every future condition.

CCaching saves energy only if exactly one subsystem requests context.

DOne sensing operation serves four users; three later reads are cheap cache hits.

Show answer

Answer: D The first reader pays the full sensing cost and populates the cache; while the value stays fresh, the remaining readers reuse it with cheap lookups.

Retrieval practice

Recall check 4 of 5

Battery Bruno says: answer from memory, then check your reasoning.

Q5A full context sense costs 0.34 mA-s and a cache hit costs 0.016 mA-s. At a cache hit rate of 80%, what is the average energy per request?

AAbout 0.081 mA-s, from 80% cheap hits and 20% full sensing.

BAbout 0.178 mA-s, averaging the two path costs without hit weights.

CAbout 0.34 mA-s, charging for full sensing on every cache hit.

DAbout 0.016 mA-s, treating every request as a cheap cache hit.

Show answer

Answer: A The blend weights the cheap lookup by the hit rate and the full sense by the miss rate: 0.0128 + 0.068 = 0.081 mA-s. That is about 4.2x cheaper than always sensing (0.34 mA-s).

Retrieval practice

Recall check 5 of 5

Battery Bruno says: answer from memory, then check your reasoning.

Q6Several components request the same context about every 5 s. An engineer shortens the cache validity window from 60 s to 2 s hoping for fresher context, and battery life gets much worse. Why?

AA shorter window makes each individual sensing operation draw more current.

BA 2 s window forces almost every request to miss and re-sense.

CShorter windows disable the accelerometer, so GPS runs constantly.

DFreshness has no effect on energy; the change must be a measurement error.

Show answer

Answer: B Freshness is the knob that sets the hit rate.

Print reference

Answers 1 of 2

Answer key.

B · ACE cache reuse is allowed when the context value is fresh, trusted, permitted, and adequate for the requesting decision.
B · Rule promotion depends on support, confidence, freshness, and consequence.
A · It learns supported relationships that may replace a direct sensor read under explicit thresholds.
D · The first reader pays the full sensing cost and populates the cache; while the value stays fresh, the remaining readers reuse it with cheap lookups.

Print reference

Answers 2 of 2

Answer key.

A · The blend weights the cheap lookup by the hit rate and the full sense by the miss rate: 0.0128 + 0.068 = 0.081 mA-s. That is about 4.2x cheaper than always sensing (0.34 mA-s).
B · Freshness is the knob that sets the hit rate.