# 4. The need for a major research thrust on risk and failure

IA et Robotique

13 nov. 2013 (il y a 8 années et 3 mois)

320 vue(s)

Risk
-
Based Control System Design

Talk Outline

1.
Introduction: the need to more deeply incorporate
failure

and
risk

into engineering

2.
Conceptual frameworks of
failure

and
risk

3.
Ideas from classical risk
-
theory

4.
The need for a major research thrust on risk and failure
sensitive control design

ALL SYSTEMS FAIL!

ALL SYSTEMS FAIL!

TYPES OF RISK IN
ENGINEERED SYSTEMS

Uncertainty
-
based risk

Information
-
based risk

Complexity
-
based risk

Model
-
based risk

TYPES OF RISK IN
ENGINEERED SYSTEMS

RISK AS AN ENGINEERING
DESIGN PARAMETER

Risk level

# of Failures

1
-
sigma

Less than 4 in 10

2
-
sigma

Less than 5 in 100

3
-
sigma

Less than 3 in 1000

4
-
sigma

Less than 7 in 100,000

5
-
sigma

Less than 6 in 10,000,000

6
-
sigma

Less than 2 in
1000,000,000

CLASSICAL MODELS OF RISK

Bernoulli trials:

Probability of failure =
p
, probability of success = 1
-
p

MTTF =

Variance =

For high
-
probability failures, MTTF is a good predictor.
For low probability of failure, it isn’t.

Associated with constraints on temporal and spatial distribution
of information

Information
-
based risk

Risk arises from loss of information, bit
-
errors, noise, component
failures, communications link failures, interconnection patterns
being poorly matched to performance objectives, etc.

Complexity
-
Based Risk in Competitive Athletics

Game dynamics have
extreme sensitivity

to:

1.
Ball speed,

2.
Ball spin,

3.
Part of racquet face striking the ball,

4.
Place at which the ball strikes the wall or
floor

Model
-
based Risk: Information Flow Must
Be Localized and Dynamically Reconfigured

Given the geometry of a formation,
how many distinct “stable
information
-
flow patterns” will
support it?

Research is needed

1.
To extend ideas from risk
-
sensitive stochastic control and
classical insurance theory to model and deal with failures in
sensor/acuator networks and multiagent robotic systems;

2.
To understand how decentralized control designs degrade as
result of changes in scale;

3.
To understand risk associated with operating under
constraints on the flow of real
-
time information;

4.
To understand risk associated with small changes in
operating conditions;

5.
To develop a theory competitive robotic athletics;

6.
To develop a formal theory of attention and information
sharing patterns for multi
-
agent systems