On Sensor Networking and Signal Processing for Smart and Safe Buildings

1
On Sensor Networking and Signal Processing for
Smart and Safe Buildings

• Pramod K. Varshney
• Department of Electrical Engineering and Computer
  Science
• Syracuse University
• 121 Link Hall
• Syracuse, New York 13244 USA

2
Overall Structure of the Center

• Strategically Targeted Academic Research
• 9 Academic Institutions
• 2 not-for-profit Research institutes

• Technology Transfer
• 50 Corporate Partners
• Fosters University/Industry collaboration

• Regional Partnership of Industry Academe
• Strategically Targeted Academic Research
• Technology Transfer and Commercialization

3
Centers Hub and Distributed Facilities
4
Outline

• Introduction
• Key challenges and issues
• Illustrative examples
• Concluding remarks

5
Indoor Air Pollution
Do you work in a Toxin Factory? Business Week
June 5, 2000
6
Societal and Economic Drivers

• Health
• 17.7 million asthma cases (4.8 million children)
• 50-100 thousand annual deaths due to elevated
  levels of particulate matter
• Productivity
• 40 to 250 billion productivity loss due to poor
  IEQ
• Sustainability
• 110 billion annual economic loss due to air
  pollution in urban areas
• 40 of total building energy consumption is for
  environmental control (over 15 of total US
  energy consumption)
• Security
• Built and urban environments are vulnerable to
  chemical/biological threats

7
The Problem

• Wide spectrum of buildings
• Residences, schools, hospitals, apartment
  buildings, office buildings, factories,
  high-valued assets
• Indoor air quality goals
• Health
• Productivity
• Exposure and risk
• Energy consumption cost
• Scenarios
• Routine day-to-day
• Health, productivity, costs
• Time to react is not critical
• Emergency
• Safety, exposure
• Rapid response required
• Affordability and cost issues
• New Buildings
• Retrofit

8
The Problem

• Some current solutions
• A single thermal sensor
• Uneven/asymmetric conditions
• inefficient
• Provide multiple knobs
• Control system is not adequate
• Replace indoor air by fresh air frequently
• Too costly
• Hybrid and demand-controlled ventilation
• Use sensing and control
• Maximize benefits of natural driving forces
• Control needed due to changing weather conditions

9
Motivation

• These and other current solutions are fairly
  primitive!
• They use one size fits all solutions and do not
  reduce human exposure and maximize comfort to the
  desirable extent
• Due to a wide spectrum of buildings and their
  scales, multiplicity of goals, and response time
  requirements, intelligent solutions are required!

10
Why Distributed Large-scale Wireless Sensor
Networks?

• Higher resolution and fidelity data available in
  a sensor-rich environment for customized
  environments
• Improved IAQ at different scales, e.g., personal
  level, thus increasing productivity without much
  increase in cost
• Rapid response in emergency situations
• Improved reliability and robustness
• More degrees of freedom for distributed control
• Enabling technologies are fairly mature for
  practical applications

11
Conceptual Process Diagram
12
Key Components

• Sensor Networks
• Topology, architecture, protocols and management
• Intelligent Information Processing
• Information fusion, learning algorithms, and
  knowledge discovery
• Control and Mitigation Methodology
• Control worthy models based on reduced order
  models, hierarchical distributed control,
  mitigation and evacuation

13
Distributed and Pervasive Sensing Paradigm
14
Challenges and Issues in i-EQS Sensor Networks
Lack of design principles for sensor networks in
buildings

• Distribution among wired and wireless sensors is
  not known
• Sensor network architecture including topology,
  number and placement of sensors, and protocols
  has not been addressed.
• Resource management including bandwidth and
  energy management has not been investigated.
• Security and information assurance requirements
  are not well understood.

Challenge 1
Challenge 2
Challenge 3
Challenge 4
15
Challenges and Issues in i-EQS Information
Processing
Lack of intelligent information processing
algorithms that fully exploit all available
information

• Inferencing and control mostly based on single
  sensor measurements.
• Systems do not take full advantage of networked
  sensors, information fusion and intelligent
  signal processing algorithms.
• Spatial and temporal dimensions (e.g.
  forecasting) are not explored in detail.
• Systems are not robust and responsive to
  evolving dynamic situations.

Challenge 1
Challenge 2
Challenge 3
Challenge 4
16
Challenges and Issues in i-EQS Control
Lack of robust multi-level intelligent
model-based control algorithms Event and state
recognition with incomplete information Complex,
non-linear and state/objective dependent
dynamics Slow system response Resources
constraints, e.g, sensors, actuators, computing
power, bandwidth
Challenge 1
Challenge 2
Challenge 3
Challenge 4
17
Sensor Placement Problem

• Problem Determining the locations where sensors
  should be placed, maximizing coverage and
  detection capability while minimizing cost
• Factors and Problem Parameters
• Building layout
• Air inlet and outlet (HVAC) locations
• Air flow simulation and analytic models
• Sensor characteristics and costs
• Approach
• Multiobjective optimization
• Modeling each candidate configuration of sensors
  as a point in a multidimensional space
• Applying evolutionary algorithms to sample search
  space effectively and efficiently

18
Data Fusion Issues

• Problems
• Detecting the presence of activities of interest,
  e.g., abnormally high pollutant concentration
• Classifying the type of activity, e.g., the type
  of pollutant
• Factors and Problem Parameters
• Sensor Characteristics in terms of their
  detection ability
• Sensor location and coverage
• Approach
• Distributed detection theory decision fusion
• Algorithms to deal with uncertainties modeling
  errors, asynchronous information
• Adaptation to changing environmental conditions

19
Decision Fusion
20
Design of Fusion Rules

• Input to the fusion center ui, i1, , N
• Output of the fusion center u0
• Fusion rule logical function with N binary
  inputs and one binary output
• Number of fusion rules 22N

21
Optimum Decision Fusion

• The optimum fusion rule that minimizes the
  probability of error is

P. K. Varshney, Distributed Detection and Data
Fusion, Springer, 1997
22
Inferencing in Distributed Sensor Networks

• Problems
• Detecting relationships between pollutant
  concentrations at different locations
• Detecting locations of abnormally high pollutant
  sources

• Factors and Problem Parameters
• Fluid flow models and simulations
• Pollutant source models and locations
• Potential sensor locations
• Approach
• Inferencing with time-sensitive probabilistic
  (Bayesian) network models

23
Illustrative Examples

• UC Berkeley study shows that the use of multiple
  sensors and ad hoc control strategies (Single
  HVAC) reduced energy consumption as well as
  predicted percentage dissatisfied (PPD)
• Energy-optimal scheme
• 17 reduction in energy consumption
• 6 reduction in PPD 30?24
• Comfort-optimal scheme
• 4 reduction in energy consumption
• 10 reduction in PDD 30?20

N. Lin, C. Federspiel and D. Auslander,
Multi-sensor Single-Actuator Control of HVAC
Systems, Int. Conf. For Enhanced Building
Operations, Richardson, TX, 2002
24
Intelligent Control of Building Environmental
Systems for Optimal Evacuation PlanningbyJ.S.
Zhang1, C.K. Mohan2, P. Varshney2, C. Isik2, K.
Mehrotra2, S. Wang1, Z. Gao1, and R. Rajagopalan
21Dept. of Mechanical, Aerospace and
Manufacturing Engineering 2Dept. of Electrical
Engineering and Computer Science
Environmental Quality Systems Center
(http//eqs.syr.edu/) College of Engineering and
Computer Science Syracuse University
25
i-BES for Optimal Evacuation Planning
Occupant
Personal Env.
Outdoor Airshed
Zone/ Room
Multizone Building
Multi-level Controls
3
2
1
0
Prediction of Pollutant Dispersion
Optimization of Peoples Movement
Monitoring of BES Conditions
Predictive control algorithm
Simulated Control Operations
26
Pollutant Dispersion in a 6-zone testbed
27
Pollutant Dispersion Multizone Model Simulations
e
Zone 6
Zone 5
b
Zone 1
e
e
Release at Outdoor Air Intake
Zone 4
Zone 3
Zone 2
a
a
Exhaust
c
e
28
Pollutant Dispersion Control and Evacuation Plan
Multizone Model Simulation Results
Concentration change over time Evacuation
routes
29
A 73-Zone Example (a floor section of 22,000 ft2)
30
Concluding Remarks

• Management of indoor air quality is an
  interesting and challenging application.
• Theory and implementation is in its infancy.
• Design of the headquarters of the Center of
  Excellence is underway. It will serve as a
  testbed for the new technology.