Ex Parte Bodenheimer et al

Patent Trial and Appeal Board

Aug 18, 2014

11331629 (P.T.A.B. Aug. 18, 2014)

UNITED STATES PATENT AND TRADEMARK OFFICE
________________

BEFORE THE PATENT TRIAL AND APPEAL BOARD
________________

Ex parte ROBERT EDWARD BODENHEIMER, JR. and
RICHARD ALAN PETERS, II
________________

Appeal 2012-005981
Application 11/331,629
Technology Center 3600
________________

Before: JOHN C. KERINS, MICHAEL L. HOELTER, and
ANNETTE R. REIMERS, Administrative Patent Judges.

HOELTER, Administrative Patent Judge.

DECISION ON APPEAL

STATEMENT OF THE CASE
This is a decision on appeal, under 35 U.S.C. § 134(a), from a
rejection of claims 1, 3‒5, 7, 8, 10, 11, 13, and 14. App. Br. 4. Claims 2, 6,
9, and 12 have been canceled. App. Br. 16‒19, Claims App’x. We have
jurisdiction under 35 U.S.C. § 6(b). We AFFIRM.

THE CLAIMED SUBJECT MATTER
The disclosed subject matter “relates to the field of intelligent
autonomous robot architectures. More specifically, the invention relates to
Appeal 2012-005981
Application 11/331,629

2

determining a trajectory through a state space of the robot using
superposition of previously learned behaviors.” Spec. para. 3. Independent
claim 1 is illustrative of the claims on appeal and is reproduced below:
1. A method for determining a new trajectory for robot
motion, the method comprising:
accessing at least one behavior from a memory, each of
the at least one behavior having an associated target;
approximating the new trajectory of robot motion based
on the at least one behavior and the associated target by a least-
mean-square algorithm; and
correcting the trajectory of robot motion based on a
distance between the associated target and a new target,
wherein the robot performs the steps of approximating
and correcting.

REFERENCES RELIED ON BY THE EXAMINER
Kasagami US 5,315,222 May 24, 1994

Li Jun and Tom Duckett, Robot Behavior Learning with a
Dynamically Adaptive RBF Network: Experiments in Offline and Online
Learning, Proc. of the 2nd Int’l Conference on Computer Intelligence,
Robotics and Autonomous Systems (CIRAS), Singapore, (December 18,
2003) (hereinafter “Jun and Duckett”).1
Li Jun and Tom Duckett, Learning Robot Behaviors with Self-
Organizing Maps and Radial Basis Function Networks, Proc. of the Second
Swedish Workshop on Autonomous Robotics, Stockholm, Sweden, 59-67
(2002) (hereinafter “Jun et al.”).2

1 Both the Examiner and Appellants identify this reference as “Jun and
Duckett.” We do the same.
2 Both the Examiner and Appellants identify this reference as “Jun et al.”
We do the same.
Appeal 2012-005981
Application 11/331,629

3

THE REJECTIONS ON APPEAL
Claims 1, 3, 4, and 143 are rejected under 35 U.S.C. § 103(a) as being
unpatentable over Jun and Duckett. Ans. 5.
Claims 1, 5, 7, 8, 10, 11, and 13 are rejected under 35 U.S.C. § 103(a)
as being unpatentable over Kasagami and Jun et al. Ans. 8.

ANALYSIS
The rejection of claims 1, 3, 4, and 14
as being unpatentable over Jun and Duckett
Appellants separately argue claim 1 and also present separate
arguments with respect to dependent claim 14. App. Br. 10, 11, 14.
Appellants also contend that claims 3 and 4, which depend directly and
indirectly, respectively, from claim 1, “are allowable for similar reasons set
forth above,” i.e., with respect to claim 1. App. Br. 12. We separately
review claims 1 and 14 with claims 3 and 4 standing or falling with claim 1.
See 37 C.F.R. § 41.37(c)(1)(vii) (2011).
Claim 1
The Examiner relies on this single reference, i.e., Jun and Duckett, for
disclosing the various steps of method claim 1 and also identifies certain
passages therein where such steps are disclosed. Ans. 6. However, the
Examiner finds that this reference is silent regarding recalling information
“from a memory” as claimed but finds that “techniques for storing
information on to a memory and later recalling the information from the

3 The Examiner initially also listed claims 5, 7, 8, 10, 11, and 13 as being
rejected over Jun and Duckett but acknowledged that their inclusion “was
the result of a typographical error” and consequently withdrew the rejection
of these claims under Jun and Duckett. Ans. 4; see also Reply Br. 2.
Appeal 2012-005981
Application 11/331,629

4

memory were very well known in the art at the time of the [A]pplicants’
invention.” Ans. 6‒7. The Examiner concludes that it would have been
obvious “to store the training information onto a memory for the purpose of
recalling the information when needed.” Ans. 7. Appellants do not dispute
the Examiner’s finding regarding the recalling of information “from a
memory.”
Appellants initially contend that “Jun and Duckett fails to disclose the
‘new target,’ as claimed.” App. Br. 10. For clarity, the limitation in
question is “correcting the trajectory of robot motion based on a distance
between the associated target and a new target” (claim 1). Paraphrasing
Appellants’ Specification, we understand that “associated target” is the
location associated with a learned (stored) behavior while a “new target” is
just that, a new location. See e.g., Spec. paras. 34 and 42 (see also Spec.
para. 32 providing guidance that “a target represents a physical location”).
The Examiner finds that this limitation is disclosed on page 3, steps 10‒12
and also page 5, second column, first two full paragraphs of Jun and
Duckett. Ans. 6. The Examiner also addresses such robot behavior as “wall
following behavior, a moving object following behavior, an obstacle
avoiding behavior and/or a path learning behavior” that is additionally
discussed in Jun and Duckett. Ans. 10‒11; Jun and Duckett, pg. 4, Section
3.1. With respect to the identified “moving object following behavior,” Jun
and Duckett describes this behavior as “the robot follows a moving object by
keeping a constant distance to it.” Jun and Duckett, Fig. 4 and Section 3.1.
As such, we understand that as the object moves relative to the robot within
the test area, a “new target” location of the object is repeatedly provided for
the robot to follow so as to maintain a constant distance with it.
Appeal 2012-005981
Application 11/331,629

5

Accordingly, we are not persuaded by Appellants’ contention that “Jun and
Duckett fails to disclose the ‘new target,’ as claimed” nor are we persuaded
by Appellants’ contention that Jun and Duckett fails to disclose “‘ . . . robot
motion based on a distance between the associated target and a new target,’
as recited in independent claim 1.” App. Br. 10.
Regarding the “correcting the trajectory of robot motion” component
of this limitation, Appellants contend that “the portion of Jun and Duckett
that the Examiner cites merely discloses a difference between Y and Ybar.”
App. Br. 10. For clarity, Jun and Duckett describes Y as a dimensional
output vector which is a function of a dimensional input vector X. Jun and
Duckett, pg. 2, second column. Likewise, Ybar is identified as a weighted
function (W) of X. Jun and Duckett, pg. 3, step 9. Nevertheless, with
respect to this limitation, the Examiner has referenced more than simply “a
difference between Y and Ybar” (which can be found at step 10) because the
Examiner also identified “page 3, steps 10-12; page 5, second column, first
two full paragraphs” of Jun and Duckett as disclosing this limitation. Ans.
6. Appellants’ appear to focus on only the single equation (i.e., the
“difference between Y and Ybar”) depicted in step 10 without considering the
other steps also identified that involve updating dimensional calculations
until a desired error range is achieved. Jun and Duckett, pg. 3, steps 10‒12.
Appellants also fail to address the teachings on page 5 of Jun and Duckett
(also identified by the Examiner, see Ans. 6) which discuss the need for
multiple iterations of a “Grab” exercise so as to diminish error when training
the robot. In short, Appellants do not persuasively explain how these steps
10 to 12 and page 5 (of Jun and Duckett) identified by the Examiner and
which go beyond “merely” ascertaining a difference (App. Br. 10) fail to
Appeal 2012-005981
Application 11/331,629

6

disclose “correcting the trajectory of robot motion” as claimed. In other
words, Appellants do not persuasively explain how the Examiner erred in
relying on Jun and Duckett for disclosing this limitation. Accordingly,
Appellants’ contention that Jun and Duckett fails to teach the step of
“correcting the trajectory of robot motion based on a distance between the
associated target and a new target” (claim 1) is not persuasive.
Appellants also address another limitation4 involving approximating
that employs a least-mean-square algorithm contending that “Jun and
Duckett teaches away from the claimed invention.” App. Br. 11.
Appellants’ contention is not persuasive because the Examiner identifies
where Jun and Duckett discusses the use of a “least-mean-square algorithm”
in the approximations made.5 Ans. 6. In fact, in making these
approximations, Jun and Duckett specifically refers to this “least mean
square (LMS) algorithm” as “well-known” (Jun and Duckett, pg. 2, col. 2,
second full paragraph) and Appellants do not explain otherwise, i.e., how its
use in approximating robot motion is not obvious. Additionally, we are
instructed by our reviewing court that a teaching away requires a reference
to actually “criticize, discredit, or otherwise discourage” investigation into
the claimed solution. In re Fulton, 391 F.3d 1195, 1201 (Fed. Cir. 2004). In

4 “[A]pproximating the new trajectory of robot motion based on the at least
one behavior and the associated target by a least-mean-square algorithm”
(claim 1).
5 Jun and Duckett discloses the use of a Radial Basis Function (RBF)
network stating that this network, trained with the well-known LMS
algorithm, possesses many good properties, including “optimal
approximation,” and that this network makes “it possible to dynamically
adapt its topological architecture for different learning tasks.” Jun and
Duckett, pg. 2, col. 2, second and third full paragraphs.
Appeal 2012-005981
Application 11/331,629

7

the present matter, Appellants do not indicate where Jun and Duckett
criticizes, discredits or otherwise discourages the use of such an algorithm
when approximating movement. In short, Appellants’ contention is not
persuasive of Examiner error and accordingly, we sustain the Examiner’s
rejection of claims 1, 3, and 4.
Claim 14
Claim 14 depends from claim 1 and includes the additional limitation
of multiplying a function by a weight matrix that is “inversely proportional”
to another matrix pertaining to radial basis function intensities. Appellants
do not dispute the teaching in Jun and Duckett of a weight matrix W or a
matrix of radial basis function intensities (i.e., “search factor τ.”). See Ans.
7; App. Br. 14. Instead, Appellants contend that “W is not inversely
proportional to the search factor τ in Jun and Duckett.” App. Br. 14. Here,
the Examiner references Jun and Duckett, pages 2 to 3, steps 9 to 11 as
disclosing this inverse relationship. Ans. 7. Step 11 provides an equation
involving a weight matrix that is based upon a value η1 (i.e., “learning rate”).
Step 11 further identifies this learning rate η1 as being inversely proportional
to search factor τ. Hence, an inverse relationship is disclosed between the
weight matrix and search factor τ, and Appellants do not persuasively
explain how this is not the case. Accordingly, we sustain the Examiner’s
rejection of claim 14.

Appeal 2012-005981
Application 11/331,629

8

The rejection of claims 1, 5, 7, 8, 10, 11, and 13
as being unpatentable over Kasagami and Jun et al.
Appellants argue claims 1, 5, 7, 8, 10, 11, and 13 together. App. Br.
12‒14. We select independent claim 1 for review with claims 5, 7, 8, 10, 11,
and 13 standing or falling with claim 1. See 37 C.F.R. § 41.37(c)(1)(vii).
Here, the Examiner primarily relies on the teachings of Kasagami for
disclosing the limitations of claim 1 but relies on Jun et al. for teaching “a
neural network for controlling robot motion.” Ans. 8‒9. In addressing
claim 1, Appellants contend that the Examiner’s reference “to the time tj+l
required for interpolation as being the ‘distance between the associated
target and a new target,’ as recited in independent claim 1” is in error. App.
Br. 12. More specifically, Appellants contend that “Kasagami et al. is
completely silent as to ‘correcting the trajectory of robot motion’ based on
tj+l.” App. Br. 13.
The Examiner states that “the office action makes no mention of ‘time
tj+l’” while continuing to find the limitation in question taught by “Fig. 12,
step S54, and equation 34” of Kasagami. Ans. 14‒15; see also Ans. 8. It
appears the Examiner’s reference to “equation 34” in Kasagami is the source
of confusion as this equation clearly pertains to time tj+l. See Kasagami, col.
13, l. 60. However, the Examiner further identifies Kasagami, columns 13
to 14; particularly, lines 4 to 8, 42 to 58, and column 14, lines 55 to 57, for
support with respect to step S54 as well as ∆u (unit displacement) illustrated
in Figure 12. Ans. 15. We note with particularity that column 13, lines 42
to 45 pertain to equation 32 and that this equation 32, along with the bulk of
the portions of Kasagami referenced by the Examiner, pertains to ∆u and not
tj+l. The Examiner thus referenced Kasagami’s unit displacement ∆u for
disclosing the limitation in question which is directed to distance and the
Appeal 2012-005981
Application 11/331,629

9

Examiner also, perhaps mistakenly, referred to an equation pertaining to
time tj+l. Appellants’ contentions only address the variable tj+l to show
Examiner error and do not address the Examiner’s more numerous
references to unit displacement as teaching this limitation. App. Br. 12‒13.
Hence, regardless of the Examiner’s reference to the tj+l set forth in equation
34, Appellants fail to address the Examiner’s findings regarding the ∆u set
forth in equation 32 and the other locations in Kasagami that likewise
pertain to unit displacement. Accordingly, Appellants have failed to respond
to the unit displacement basis of the Examiner’s rejection and as such,
Appellants have not shown how the Examiner’s findings based on ∆u are in
error.
Appellants also contend that “[t]he algorithm applied in the claimed
invention differs significantly. The claimed robot is an articulated dexterous
manipulator, not a mobile platform.” Reply Br. 3. However, no particular
algorithm is recited in claim 1 (other than the well-known LMS algorithm,
see supra), and further, Appellants do not indicate any limitation that
distinguishes between the type of robot involved. Additionally, Appellants
contend that the steps 10 to 12 previously referenced adjust “the RBF net for
a completely different purpose than does the claimed system.” Reply Br. 4.
However, the preamble of claim 1 recites “[a] method for determining a new
trajectory for robot motion” and Appellants do not explain how the
references relied upon by the Examiner fail to also address robot motion.
In view of the record presented, we sustain the Examiner’s rejection
of claims 1, 5, 7, 8, 10, 11, and 13.

Appeal 2012-005981
Application 11/331,629

10

DECISION
The Examiner’s rejections of claims 1, 3‒5, 7, 8, 10, 11, 13, and 14
are affirmed.
No time period for taking any subsequent action in connection with
this appeal may be extended under 37 C.F.R. § 1.136(a)(1)(iv).

AFFIRMED

llw