Cyber-Physical Systems

Apprenticeship Learning for Cyber-Physical System Intelligence  ◾  59

© 2010 Taylor & Francis Group, LLC

min

wx,

ω

2

2

such that

µµ

=≥

∑

xx

i

i

i

()

;0

x

i

i

=≥

∑

10

; ω

ω ≥ μ − μ

E

; ω ∈ W

ese formulae limit the weight to a nonnegative value. Figures 5.2 through 5.4 show the learning

effects from an expert’s demo.

Figure 5.3 The left driving simulation shows the white car exhibiting a nice driving style and

staying away from traffic and in the left lane. The right simulation shows the white car exhibit-

ing the nasty driving style and crashing into a car in its path while surrounded by other cars.

Figure 5.4 Car obstacle detection. The car must be able to detec ...