An Optimum Braking Strategy

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Meccanica (2006) 41:693–696

DOI 10.1007/s11012-006-9011-0

BRIEF NOTE

An optimum braking strategy


R. J. Knops · Piero Villaggio

Received: 13 May 2005 / Accepted: 4 June 2006 / Published online: 31 October 2006
© Springer Science+Business Media B.V. 2006

Keywords Rigid body dynamics · Friction which, of course, cannot be exceeded. Accordingly,
it is of interest to devise an optimum strategy for
braking subject to these constraints.
1 Formulation of the problem In mathematical terms, the problem can be for-
mulated as follows. Consider a material point of
Car-drivers often are required to bring their vehicle mass M moving in a straight line with variable
to a sudden stop when at a given speed to avoid speed v(t) and subject to a resistance force R
colliding, for example, with a fixed obstacle (see assumed to be a function of time and speed. At
Fig. 1). The first impulse is to exert and maintain some initial instant t = 0 the speed has a given
maximum pressure on the brake until the car is value v(0) = v0 , and at a later instant T > 0 the
halted, but this response is inadvisable since exces- speed must be zero so that v(T) = 0. The distance
sive deceleration may cause skidding, and also may traversed by the mass in first coming to rest is
 T
result in injury to the driver and passengers even
L= v(t) dt , (1.1)
when seat belts are worn. Just as importantly, a con- 0
stant thrust on the brake pedal does not produce where the speed is the solution to the equation of
constant deceleration. Decades of experimental motion
research on how shoe brakes effect the motion
of wheels have conclusively demonstrated that the M v̇ + R(v, t) = 0 (1.2)
magnitude of the braking force decays rapidly with that satisfies the end conditions v(0) = v0 , v(T) = 0.
speed. It follows that, when the driver brakes at Equation 1.2 is the generalisation of all laws of
high initial speed, the braking force, small at the braking proposed in the last 50 years. It includes a
beginning, gradually increases to a given amount tribological law due to Kragelskii [2, p. 170–180]
for the dry braking of a jaw on the wheel of a
R. J. Knops (B)
Department of Mathematics and the Maxwell Institute
for Mathematical Sciences, Heriot-Watt University,
EH14 4AS, Edinburgh, UK
e-mail: R.J.Knops@ma.hw.ac.uk

P. Villaggio
Dipartimento Ingegneria Strutturale, Università di
Pisa, via Diotisalvi, 2, 56126 Pisa, Italy Fig. 1 Schematic representation
694 Meccanica (2006) 41:693–696

machine, and is also applicable to vehicles equipped Levi-Civita and Amaldi [3, Vol. 1, Ch. 9], assume
with rubber tyres in which the slip between tyre and that the braking force R is constant, and conse-
road depends on the speed (cf. [1, p. 103]). Even quently that the length L and time T are deter-
recent antilock braking systems (ABS) can be de- mined solely by the initial conditions.
scribed by (1.2) provided that the surface of the
road is homogeneous.
Braking is also affected by the rolling resistance 2 The variational solution
and the aerodynamic drag, but their influence,
important in the waste of fuel during the whole Since R(v, t) has the form (1.3), where 0 ≤ f (t) ≤ 1
run, is negligible with respect to the longitudinal and g(v) ≥ 1, we immediately deduce from (1.2)
force exerted by the road on tyres during braking that
(cf. [1, p. 72, 134]. Another effect disregarded in v̇(t) ≤ 0 , for 0 ≤ t ≤ T, (2.1)
(1.2) is the unavoidable delay existing between the
instantaneous action of braking and application of and therefore
the force R(v, t). However, the neglect is justified 0 ≤ v(t2 ) ≤ v(t1 ) ≤ v0 , 0 ≤ t1 ≤ t2 ≤ T . (2.2)
by the fact that the delay (about 0.02 s) is very small
compared to the average brake time (about 10 s). Set
 t
In (1.2), the force R(v, t) may be further speci-
F(t) = f (τ )dτ , (2.3)
fied by taking account of the experimental 0
 v
results already mentioned, and a particular exam-
G(v) = g(ν)dν , (2.4)
ple is considered in Section 3 devoted to a numeri- v0
cal illustration. Consequently, the general form for
and integrate (1.2) to obtain
R(v, t) may be represented by
G(v) + F(t) = 0 . (2.5)
f (t)
R=M , (1.3) But from (2.3) and the properties of f (t) we con-
g(v)
clude that
dF(t)
where f (t) (0 ≤ f (t) ≤ 1) is a control function F(t) ≥ 0 , ≥ 0, 0 ≤ t ≤ T , (2.6)
intended to reflect the characteristic increase or dt
relaxation of the brake pressure spontaneously while (2.5) and the properties of g(v) imply that
decided by the driver, and g(v), which has dimen- dG(v)
sion inverse to that of acceleration, is a strictly G(v) ≤ 0 , ≥ 0, 0 ≤ t ≤ T , (2.7)
dv
increasing function of speed (i.e. ġ(v) > 0), and,
so that G(v) increases with v.
for suitable choice of units, is supposed to satisfy
Let f1 (t), f2 (t) be different functions such that
the condition g(0) = 1. These assumptions imply
that the braking force is inversely proportional to f1 (t) ≤ f2 (t) , 0≤t≤T,
an increasing function of the speed and tends to and let v1 (t), v2 (t) be corresponding solutions to
zero as v tends to infinity. (2.5). Then, (2.3) and (2.5) give
This formulation of the optimum braking strat-
egy can be recognised as a standard problem in − G(v1 ) = F1 (t) ≤ F2 (t) = −G(v2 ) , (2.8)
optimum control (cf. [4, Sec. 32]) in which the con- which together with (2.7) lead to
trol is provided by the function f (t) that minimises
the functional L given by (1.1) from among all func- v2 (t) ≤ v1 (t) .
tions v(t) satisfying equation (1.2), the end condi- A minimum for L thus is obtained on selecting the
tions v(0) = v0 , v(T) = 0, and whose time rate maximum value for f (t), namely
of change (deceleration) possesses a prescribed
f (t) = 1 , 0≤t≤T, (2.9)
upper bound a0 (v̇(t) ≤ −a0 ).
We further comment that classical textbooks de- whenever this is possible. But (1.2) with (1.3) and
voted to Rational Mechanics, for example that by the choice (2.9) indicate that v̇(t) decreases as
Meccanica (2006) 41:693–696 695

−1/g(v) and hence there is a time tc , 0 ≤ tc ≤ T, at Finally, an expression for the stopping distance
which the acceleration assumes the value L is derived as follows:
 T  tc  T
v̇(tc ) = −a0 . (2.10)
L(f ) = v(τ )dτ = v(τ )dτ + v(τ )dτ
0 0 tc
Of course, there are functions g(v) for which tc  tc 1
may neither exist nor be equal to either 0 or T. = a0 (T − tc )2
v(τ )dτ +
As an example of such possibilities, assume that 0 2
 tc
tc = 0, T, and that (2.10) holds. From (1.2) we 1
= G−1 (−τ )dτ + a0 (T −tc )2 ,
have g(v(tc )) = 1/a0 , and when this equation pos- 0 2
sesses a solution v(tc ) ≥ v0 there is a contradiction. (2.21)
Accordingly, we suppose the functions g(v) consid-
ered here are always consistent with the assump- where tc is given by (2.19).
tion that 0 < tc < T, so that during the remaining
period (tc , T) we must have
3 Numerical example
v̇(t) = −a0 , tc ≤ t ≤ T , (2.11)
which after integration leads to To illustrate our results, we consider a particular
representation for g(v) proposed by Kragelskii and
v(t) = a0 (T − t) , tc ≤ t ≤ T . (2.12)
based upon his cited experimental data [op. cit.
From (1.2) the function f (t) is given by p. 179] but slightly modified to conform with the
dimensions stipulated by our form (1.3) for the
− a0 g(v(t)) + f (t) = 0 , tc ≤ t ≤ T , (2.13)
force R:
and we conclude that
g1 (v) = (M/µ0 P)(1 + α|v|) , (3.1)
f (t) = a0 g(a0 (T − t)) , tc ≤ t ≤ T . (2.14)
where µ0 is the coefficient of friction between rub-
On the other hand, (2.5) becomes ber and asphalt, P is the weight of the mass M,
G(v(t)) + t = 0 , 0 ≤ t ≤ tc , (2.15) and α is a non-negative constant (α ≥ 0) whose
mean value, obtained from experiment, is given by
and by continuity it follows that α = 0.03 ms−1 , when the car’s speed is measured
G(vc ) + tc = 0 , (2.16) in meters per second. The expression (3.1) for g(v)
does not satisfy the condition g(0) = 1, and rescal-
where we have introduced the definition:
ing of distance is required. Let x = β y, where
vc = a0 (T − tc ) . (2.17) β = µ0 P/M so that
Again, (2.13) evaluated at t = tc yields v(t) = dx/dt = β dy/dt = βV(t) , (3.2)
a0 g(vc ) = 1 , (2.18) and dv/dt = β dV/dt.
and, on recalling (2.1) and the assumptions on g(v), The equation of motion (1.2) becomes
we deduce that V̇ + f (t)/g(V) = 0 , (3.3)
−1
a0 (T − tc ) = g (1/a0 ) , (2.19) where the function g(V) is given by
which enables us to rewrite (2.16) as g(V) = (1 + α β V) . (3.4)
G(g −1
(1/a0 )) = a−1
0 g
−1
(1/a0 ) − T , The developments of Section 2 now apply with v(t)
or replaced by V(t) given by (3.2), and the maximum
 g−1 (1/a0 ) permissible deceleration rescaled to be b0 = a0 /β.
g(ν) dν = a−1
0 g
−1
(1/a0 ) − T . (2.20) We set µ0 = 0.6 (cf. [5, p. 309]), while the ratio
v0 P/M, by definition the acceleration due to grav-
Notice that, because a0 , T, and g are prescribed, ity, is given by 9.81 m s−2 and so β = 9.81 × µ0 =
relation (2.20) determines the value of v0 . 5.886. Moreover, we suppose that the stopping time
696 Meccanica (2006) 41:693–696

is selected to be T = 15 s, and that the maximum Finally, the total stopping distance L is deter-
permissible deceleration is a0 = 3 m s−2 . mined from (2.21) which from (2.4) and (2.15)
Expression (2.18), when inverted and adjusted yields L = 335.97 m. Of this amount, 164.57 m is
for rescaling, gives the value travelled at maximum brake pressure.
It is interesting to observe that, on applying
g−1 (1/b0 ) = (1/α β) (1/b0 − 1) , (3.5)
the standard theory of braking according to which
and (2.20) becomes g(v) = M/(µ0 P) and the deceleration is not con-
 g−1 (1/b0 ) strained, the car will come to rest after 7.64 s within
g(τ )dτ = b−1
0 g
−1
(1/b0 ) − T . (3.6) a distance L = 163.78 m.
V0

We may solve (3.6) for V0 after appeal to (3.4).


The subsequent conversion to the original variable References
gives v0 = 43.93 m s−1 for the stated numerical
1. Dixon JC (1991) Tyres, suspesion and handling. Cam-
data. At greater initial speeds the car cannot be bridge University Press, Cambridge
stopped in the given time subject to the assumed 2. Kragelskii IV (1965) Friction and wear. Butterworths,
constraints. London
To determine the time tc we substitute g(V) from 3. Levi Civita T, Amaldi V (1950) Lezioni di meccanica
razionale. Zanichelli, Bologna
(3.4) in (2.19) to obtain tc = 4.31 s. Consequently, 4. Petrov IP (1968) Variational methods in optimum con-
in our particular example, the driver after 4.31 sec- trol theory, Academic Press, New York, London
onds from starting to brake must relax the brake 5. Szabó I (1963) Einführung in die technische mechanik,
pressure to avoid exceeding the maximum permit- Springer, Berlin/Göttingen/Heidelberg
ted deceleration.

You might also like