The key difference between first-generation Smart Home and true SmartInterior is that first-generation systems follow rules set at installation (Rule-Based) and must be manually updated when behavior changes. SmartInterior with Behavior-Adaptive AI learns real behavior and adjusts automatically.
Rule-Based vs Behavior-Adaptive: The Critical Difference
Rule-Based System: - Rule set: turn off AC at 22:00 every night - If behavior changes (night shift work begins): rule must be manually updated - If not updated: system malfunctions every night until corrected Behavior-Adaptive System: - Observe: in the first week of night shift work, the system notices AC is turned back on at midnight every night - Learn: after 5–7 days, the system forms Weeknight Pattern B (AC stays on until 02:00) - Adapt: proposes Weeknight Pattern B for user Confirmation, or automatically applies it once Confidence reaches 85%
Three Levels of Adaptive Learning
Level 1: Preference Learning (Weeks 1–4) The system learns actual preferred Setpoints, not the values configured at installation: - Observe: user adjusts temperature from 24°C to 25°C every time before sleeping → automatically updates Bedtime Setpoint to 25°C - Observe: user turns kitchen lights to full 100% while cooking but dims to 30% when eating → creates separate Cooking Mode and Dining Mode Level 2: Schedule Learning (Months 1–3) The system learns differences between days: - Weekday Pattern versus Weekend Pattern - Special day patterns (the one day per week with a morning workout) - Seasonal variation (summer months, AC starts 1 hour earlier) Level 3: Contextual Learning (Months 3–6) The system learns complex contextual patterns: - When AQI exceeds 150, occupants tend to stay home all day → switch HVAC to Full-Day Occupancy mode - When it rains, occupants often open windows even with AC on → learn that Rain + temperature below 28°C = Natural Ventilation Mode - When guests arrive, the system detects above-normal Occupancy → increases AC and lighting automatically
Per-Person Personalization: Multi-Profile Learning
In multi-person households, the system builds separate Behavioral Profiles: - Identity recognition: via Smartphone Presence (Bluetooth/WiFi), Biometrics (if Smart Lock present), or Behavioral Pattern alone - Separate Profiles: one occupant prefers 24°C, another prefers 26°C → shared living room negotiates 25°C Setpoint - When one person is alone in a room → that person’s Personal Preference is applied
Boundaries of Adaptive Learning: When the System Should Not Self-Adjust
A good Behavior-Adaptive System knows what it should not auto-adjust: - Safety-Critical Rules: emergency lighting, Smoke Detector Alerts are never Learned Away - Energy Budget Hard Limit: if a THB 3,000/month Hard Cap is set, the system will not Adjust above it even if it Learns the user prefers a colder setting - Security Rules: security configurations are never Auto-Adapted The system notifies via LINE OA when it detects an interesting new Pattern, allowing the user to Confirm or Dismiss before Auto-Apply.