Cooperative Artefacts: Assessing Real World Situations with Embedded Technology

1
Cooperative ArtefactsAssessing Real World
Situations with Embedded Technology

• Martin Strohbach, Hans Gellersen, Gerd Kortuem,
  Christian Kray
• Lancaster University

2
Handling and Storage of Chemicals
!
3
Incompatible Chemicals Hazard

• Incompatible materials like Calcium Hypochlorite
  and Methyl carbitol may not be stored together

4
Support Staff Dealing with Chemicals

• Current Situation
• Training Seminars and Safety Manuals
• Unreliable
• In situ assistance
• Alert potential hazards where and when they arise

HAZARD !
5
Cooperative Artefacts
6
Cooperative Reasoning

• Each container
• Autonomous Observations
• Individual assessment of situations

Ask KB
Reply
Query
Query
Hazard ?
Query
Yes/No.
Ask KB
Reply
Reply
7
Methodology

• Take concrete scenario from industry
• 3 different hazards
• Incompatible Materials
• Critical Mass
• Unapproved Area
• Build prototypes
• Testbed

8
Chemical Container

• Domain Knowledge
• reactive(ltchemicalgt,ltchemicalgt)
• critical_mass(ltchemicalgt,ltnumbergt)
• critical_time(ltchemicalgt,lttimegt)
• content(me,ltchemicalgt)
• mass(me,ltnumbergt)
• Observational Knowledge
• proximity(ltcontainergt,ltcontainergt)
• no_proximity
• location(ltcontainergt,ltin/outgt,lttimegt)

9
Scenario
10
Critical Mass Hazard
hazard_critical_mass- content(me, CH),
cond_sum(M1, (proximity(me,C),
content(C,CH), mass(C,M1)),S), mass(me, M2),
sum(S, M2, SUM) critical_mass(CH, MASS), MASS lt
SUM.
CaclciumHypochlorite
content ?
Mass?
5kg
11
Incompatible Materials Hazard
hazard_incompatible- content(me,
CH1), proximity(me, C), content(C,
CH2), reactive(CH1, CH2).
12
Unapproved Area Warning
warning- location(me, out, T1),
content(me,CH), critical_time(CH,T2), T1ltT2.
13
Unapproved Area Hazard
After one hour
hazard_unapproved- content(me, CH),
critical_time(CH, T1), location(me, out, T2),
T1 lt T2.
14
Hardware Implementation

• 2 PCBs logically linked
• US Distance Measurements
• KB, Inference Engine and LEDs
• Based on Particle Smart-its
• PIC18F6720 _at_ 20 MHz (5 MIPS)
• Program Memory 128KB
• Data Memory 3.8KB RAM, 1KB internal EEPROM
• External Memory 32 KB EEPROM

15
Inference Engine

• Based on First Order Horn logic
• Backward chaining algorithm with depth first
  search
• Restrictions
• Backtracking only over local predicates
• Restrictions in number of arguments
• Memory Footprint 37KB ROM (29), 1.9KB RAM
  (51) worst case

16
Rule evaluation

• Observations change KB
• Changes trigger re-evaluation of actuator rules
• Side effect during evaluation of actuator rules
  changes states of actuators

17
Indicating Hazards

• Hazard
• Warning
• No Hazard

18
Results

• Detection of complex situations in the physical
  world is possible
• Feasible to embed knowledge, perception and
  reasoning in artefacts
• Sufficient Response Time
• 1-2 seconds

19
Properties of Cooperative Artefacts

• No external infrastructure
• Rules of the application domain can be directly
  mapped to rules in artefacts

20
Future Work

• How to provide feedback ?
• How reasoning be limited to relevant knowledge ?
• How can hazards be detected that involve hundreds
  of containers ?

21
Thank you

• Questions ?