IN THE UNITED STATES PATENT AND TRADEMARK OFFICE
`
`___________________
`
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`___________________
`
`
`PETROLEUM GEO-SERVICES INC.
`Petitioner
`
`v.
`
`WESTERNGECO LLC
`Patent Owner
`___________________
`
`Case No. IPR2014-014771
`U.S. Patent No. 7,080,607
`
`
`
`
`
`PETITIONER PETROLEUM GEO-SERVICES INC.’S REPLY
`
`
`David I. Berl, Reg. No. 72,751
`Jessamyn S. Berniker, Reg. No. 72,328
`Thomas S. Fletcher, Reg. No. 72,383
`Christopher A. Suarez, Reg. No. 72,553
`WILLIAMS & CONNOLLY LLP
`725 12th St., NW
`Washington, DC 20005
`Telephone:
`202-434-5000
`Fax:
`202-434-5029
`
`Counsel for Petitioner, Petroleum Geo-Services Inc.
`
`
`1 Case IPR2014-00688 is a related proceeding.
`
`
`
`

`
`
`
`TABLE OF CONTENTS
`
`I.
`
`CLAIM CONSTRUCTION ............................................................................ 2
`
`A.
`
`B.
`
`C.
`
`The Board’s Interpretation of “Predicting Positions” Is Correct. ......... 3
`
`“Calculating Desired Changes” in Position Cannot Require
`Calculating Forces or a Particular Method of Calculating
`Forces. ................................................................................................... 8
`
`The Meaning of “Global Control System” Is the Same Across
`the ’607 and ’967 Patents. ..................................................................... 9
`
`II.
`
`THE GIKAS ARTICLE (EX. 1006) ............................................................. 10
`
`A. Gikas Uses a Dynamic (But Not “Behavior-Predictive”) Model
`to Estimate the Location of Any Point in the Streamer Array. ........... 10
`
`B.
`
`Gikas Discloses a Kalman Filter that Operates in Real-Time. ........... 12
`
`III. THE CHALLENGED CLAIMS ARE OBVIOUS ....................................... 14
`
`A. Modifying the ’636 PCT’s System To Use Gikas’ Kalman
`Filter Would Have Been Obvious, Making Claim 15 Obvious. ......... 15
`
`B.
`
`The Dependent Claims Would Have Been Obvious. .......................... 17
`
`1.
`
`2.
`
`3.
`
`Claims 16, 17, 19, 20, 21 and 23 Are Obvious. ........................ 18
`
`Claim 18 Is Obvious. ................................................................ 19
`
`Claim 22 Is Obvious. ................................................................ 21
`
`IV. PURPORTED “SECONDARY CONSIDERATIONS” ............................... 22
`
`V.
`
`THIS IPR IS NOT TIME-BARRED ............................................................. 23
`
`
`
`i
`
`

`
`
`
`TABLE OF AUTHORITIES
`
`FEDERAL CASES
`ABT Sys., LLC v. Emerson Elec. Co., -- F.3d --, 2015 WL 4924160
`(Fed. Cir. Aug. 19, 2015) .............................................................................. 22, 23
`
`Amkor Tech., Inc. v. Tessera, Inc., IPR2013-00242, Paper 98 (Jan. 31,
`2014) ............................................................................................................. 24, 25
`
`Apple Inc. v. Motorola, Inc., 757 F.3d 1286 (Fed. Cir. 2014) ................................... 6
`
`Aruze Gaming Macau, Ltd. v. Mgt Gaming, Inc., IPR2014-01288,
`Paper 13 (Feb. 20, 2015) ..................................................................................... 25
`
`Belden Inc. v. Berk-Tek LLC, 610 F. App’x 997 (Fed. Cir. 2015) ............................ 9
`
`Corning Glass Works v. Sumitomo Elec. U.S.A., Inc., 868 F.2d 1251
`(Fed. Cir. 1989) ..................................................................................................... 8
`
`GE Lighting Solutions, LLC v. AgiLight, Inc., 750 F.3d 1304 (Fed.
`Cir. 2014) .............................................................................................................. 6
`
`Gonzalez v. Banco Cent. Corp., 27 F.3d 751 (1st Cir. 1994) .................................. 25
`
`Hewlett-Packard Co. v. MCM Portfolio, LLC, IPR2013-00217, Paper
`10 (Sept. 10, 2013) .............................................................................................. 25
`
`In re Bigio, 381 F.3d 1320 (Fed. Cir. 2004) ................................................ 6, 7, 8, 20
`
`In re Cuozzo Speed Techs., LLC, 793 F.3d 1268 (Fed. Cir. 2015) ...................... 4, 25
`
`In re Woodruff, 919 F.2d 1575 (Fed. Cir. 1990) ................................................ 21, 22
`
`Kemco Sales, Inc. v. Control Papers Co., 208 F.3d 1352 (Fed. Cir.
`2000) ................................................................................................................... 21
`
`Linear Tech. Corp. v. ITC, 566 F.3d 1049 (Fed. Cir. 2009) ...................................... 8
`
`Motorola Mobility LLC v. Arnouse, IPR2013-00010 (Jan. 30, 2013) ..................... 25
`
`NTP Inc. v. RIM, Ltd., 418 F.3d 1282 (Fed. Cir. 2005) ............................................. 9
`
`ii
`
`

`
`
`
`Pacing Techs., LLC v. Garmin Int’l, Inc., 778 F.3d 1021 (Fed. Cir.
`2015) ..................................................................................................................... 7
`
`Phillips v. AWH Corp., 415 F.3d 1303 (Fed. Cir. 2005) (en banc) ................... 4, 8, 9
`
`Shamrock Techs., Inc. v. Med. Sterilization, Inc., 903 F.2d 789 (Fed.
`Cir. 1990) ............................................................................................................ 25
`
`Taylor v. Sturgell, 533 U.S. 880 (2008) ................................................................... 24
`
`U.S. v. Sine, 493 F.3d 1021 (9th Cir. 2007) ............................................................. 23
`
`W. Union Co. v. MoneyGram Payment Sys., Inc., 626 F.3d 1361 (Fed.
`Cir. 2010) ............................................................................................................ 18
`
`Zerto, Inc. v. EMC Corp., IPR2014-01254, Paper 35 (Mar. 3, 2015) ..................... 24
`
`
`
`OTHER AUTHORITIES
`
`35 U.S.C. § 315 (2012) ..................................................................................... 24, 25
`
`154 Cong. Rec. S9987 (Sept. 27, 2008) ................................................................... 25
`
`Trial Practice Guide, 77 Fed. Reg. 48,759 (2012) ................................................... 24
`
`iii
`
`

`
`
`
`As in the first ’607 IPR (IPR2014-00688), WesternGeco (WG) has no
`
`response to PGS’s validity challenge to the ’607 patent except to propose a claim
`
`construction of “prediction” that is inconsistent with (1) the claims, (2) the
`
`specification, (3) WG’s prior litigation positions, and, quite alarmingly, (4) the
`
`position WG advocated to the European Patent Office (“EPO”) in a brief only two
`
`weeks ago. WG apparently believes it may rewrite its claims with impunity, in
`
`direct contravention of the intrinsic evidence and its own repeated statements to
`
`multiple tribunals regarding the meaning of that intrinsic evidence.
`
`WG urges that the only “reasonable” construction of “predict positions”
`
`requires the use of a so-called “behavior-predictive model” even though PGS, Dr.
`
`Evans, the Board (preliminarily, in the Institution Decision), WG’s contentions in
`
`the ION trial, and WG’s European lawyers all agree that the specification does not
`
`require any particular prediction methodology. Only WG’s current U.S. lawyers
`
`disagree, as not even Dr. Triantafyllou, WG’s expert, was willing to adopt this
`
`position. He testified that the specification identifies as “preferred” various
`
`methods of prediction that are not behavior-predictive models, Ex. 1001 at 4:28-
`
`29; Ex. 1117 at 204-05; Ex. 1105, meaning that WG’s proposed construction reads
`
`out preferred embodiments—the antithesis of the “broadest reasonable
`
`interpretation” applicable in this proceeding. Even Dr. Triantafyllou recognized
`
`that the Board’s broader preliminary construction is not unreasonable—testimony
`
`1
`
`

`
`
`
`he provided no fewer than three times at his recent deposition. Ex. 1118 at 545,
`
`547, 548. Dr. Triantafyllou’s candor, and the disclosure of the specification that he
`
`(and WG’s prior attorneys) could not avoid, confirm the correctness of the Board’s
`
`construction and preclude adoption of the narrower construction WG advances.
`
`As in the first ’607 IPR, WG does not seriously dispute that the claims are
`
`invalid under the Board’s construction. And while in this case, because a series of
`
`dependent claims are at issue, the key references have changed—the ’636 PCT
`
`combined with the Gikas paper replaces the Workman patent—the merits have not.
`
`Applying the Board’s constructions, the experts agree that the Gikas paper teaches
`
`a Kalman filter that estimates the actual location of any point in the streamer array
`
`using prior measurements. And neither Dr. Triantafyllou nor WG dispute that the
`
`POSA would have been motivated to improve the ’636 PCT’s control system by
`
`incorporating Gikas’ Kalman filter and would have had a reasonable expectation of
`
`success in so doing. Nor does WG or its expert provide any basis to conclude that
`
`the dependent claims add any patentable distinctions. These limitations are trivial,
`
`and the dependent claims 16-23 therefore were obvious and should be cancelled.
`
`I.
`
`CLAIM CONSTRUCTION
`
`The claims at issue depend from Claim 15, which reads, in its entirety:
`
`An array of seismic streamers towed by a towing vessel comprising:
`(a) a plurality of streamer positioning devices on or inline with each
`streamer;
`
`2
`
`

`
`
`
`(b) a prediction unit adapted to predict positions of at least some of
`the streamer positioning devices; and
`(c) a control unit adapted to use the predicted positions to calculate
`desired changes in positions of one or more of the streamer
`positioning devices.
`
`Ex. 1001 at 12:26-34. As in the first ’607 IPR, the parties dispute the constructions
`
`of “predict positions” and “calculating desired changes in positions.” The
`
`additional record in this proceeding confirms the Board’s and PGS’s constructions.
`
`A. The Board’s Interpretation of “Predicting Positions” Is Correct.
`The principal dispute between the parties concerns the “predicting positions”
`
`limitation of clause (b), which the Board construed in the Institution Decision to
`
`require that the prediction unit be adapted to “estimate[] the actual locations” of at
`
`least some of the streamer positioning devices (“SPDs”). Paper 18 at 10. In
`
`adopting this construction, the Board cited the ’607 patent’s disclosure that, due to
`
`the time delays inherent in measuring positions, “the global control system 22 runs
`
`position predictor software to estimate the actual locations of each of the birds 18.”
`
`Id. (citing Ex. 1001 at 4:51-55). In other words, the prediction unit uses “old”
`
`location information to estimate current, “actual locations.”
`
`Faced with prior art systems that performed such predictions, WG and its
`
`expert retreat to five words in the specification (“behavior-predictive model-based
`
`control”) to try to rewrite the claims to require this purportedly novel feature. WG
`
`3
`
`

`
`
`
`thereby argues that the term “predict positions” requires the use of a “behavior-
`
`predictive model” that creates a physical model of the streamers based on a
`
`panoply of forces nowhere mentioned in the specification. Ex. 1103 at 165-70.
`
`The Board’s construction remains correct for at least three reasons.
`
`1.
`
`The Board’s construction is broader than WG’s construction, and
`
`WG’s own expert, Dr. Triantafyllou, agrees that it is not unreasonable. He
`
`recognized the difference between the constructions, noting that “estimation is a
`
`very general term” while his construction required a specific methodology. Ex.
`
`1118 at 545-46. Contrary to WG’s position, however, he testified no fewer than
`
`three times that the Board’s construction is “not unreasonable.” See id. at 545,
`
`547, 548. Because the Board is required to apply the broadest reasonable
`
`interpretation of the claims, In re Cuozzo Speed Techs., LLC, 793 F.3d 1268, 1275-
`
`79 (Fed. Cir. 2015), Dr. Triantafyllou’s candid acknowledgement of the
`
`reasonableness of the Board’s construction is determinative.
`
`2.
`
`The claim language supports the Board’s construction, a paramount
`
`consideration in construing a claim. See Phillips v. AWH Corp., 415 F.3d 1303,
`
`1314-15 (Fed. Cir. 2005) (en banc). Claim 15 requires a “prediction unit adapted
`
`to predict positions of at least some of the streamer positioning devices.” While
`
`the claim limits what must be predicted (at least two positions), it imposes no limit
`
`on how those positions are predicted, indicating—particularly under the applicable
`
`4
`
`

`
`
`
`BRI standard—that it covers any technique for doing so. WG continues to offer no
`
`explanation for how the language of claim 15 can be read to require a specific
`
`method— a “behavior-predictive model”—to predict positions. See POR at 5-10.
`
`3.
`
`The Board’s construction is also consistent with the specification,
`
`while WG’s construction both excludes preferred embodiments and contradicts its
`
`representations to the EPO about the specification’s disclosure.
`
`The ’607 patent discloses that its system “preferably maintains a dynamic
`
`model of each of the seismic streamers 12.” Ex. 1001 at 4:28-29. As Dr.
`
`Triantafyllou acknowledges, a “behavior-predictive model” is a subset of
`
`“dynamic” models (which themselves are only “preferably” used), and thus only is
`
`one embodiment of models for “predicting positions” disclosed in the
`
`specification. Ex. 1117 at 201, 204-05. WG’s expert made this point with
`
`exceptional clarity, drawing a Venn diagram showing that “behavior-predictive”
`
`models are a subset of “dynamic” models. Id. at 204-05 & Ex. 1105.
`
`By indicating that dynamic models are only “preferred,” the specification
`
`conveys that it contemplates other mechanisms for predicting streamer positions.
`
`WG repeatedly has interpreted the specification this way, arguing to the EPO that a
`
`“behaviour-predictive” model was just one of many ways of implementing the
`
`patent. See Ex. 1127 at ¶¶ 4.5-4.12; Ex. 1125 at 10. As WG emphasized barely
`
`two weeks ago, because the patent’s system “preferably” uses a “dynamic model,”
`
`5
`
`

`
`
`
`the specification makes “a positive disclosure that a dynamic model, whether
`
`behaviour-predictive or otherwise, is not essential.” Ex. 1125 at 10 (emphasis
`
`added); see Ex. 2073 at 2, 7 (a “predicted position determination is preferably
`
`based upon the behavior of the streamers” (emphasis added)); Ex. 1001 at 4:28-
`
`34). Thus, when not addressing this tribunal, WG trumpets that the specification
`
`discloses predicting positions using (1) behavior-predictive models (the center of
`
`Dr. Triantafyllou’s diagram, Ex. 1105); (2) other dynamic models (the broader set
`
`in Ex. 1105); and (3) other, non-dynamic models as well.
`
`WG’s factual characterization of its specification is highly relevant to claim
`
`construction, Apple Inc. v. Motorola, Inc., 757 F.3d 1286, 1312-13 (Fed. Cir.
`
`2014), and highlights the flaws in WG’s proposed construction here. WG’s
`
`construction excludes dynamic but non-behavior-predictive models, even though
`
`the specification teaches that a dynamic model of any kind is “preferred.” Ex.
`
`1001 at 4:28-29. Even under the more restrictive Phillips rubric, courts “normally
`
`do not construe claims in a manner that would exclude the preferred embodiment.”
`
`GE Lighting Solutions, LLC v. AgiLight, Inc., 750 F.3d 1304, 1311 (Fed. Cir.
`
`2014). Under the broadest reasonable interpretation standard applied here,
`
`“[a]bsent claim language carrying a narrow meaning, the PTO should only limit
`
`the claim based on the specification or prosecution history when those sources
`
`expressly disclaim the broader definition.” In re Bigio, 381 F.3d 1320, 1325 (Fed.
`
`6
`
`

`
`
`
`Cir. 2004) (emphasis added). A disclosure that dynamic models (whether
`
`behavior-predictive or not) are “preferred” is the antithesis of language that
`
`expressly disclaims them or alternative, non-preferred embodiments.
`
`WG wrongly suggests that the language “inventive control system,” which
`
`appears in the only sentence discussing the behavior-predictive model, see Ex.
`
`1001 at 4:10-14, is an express disclaimer of claim scope. Once again, WG’s own
`
`characterization of its specification refutes any such notion: citing two of the same
`
`cases on which it now relies, Honeywell and Verizon, POR at 10, WG persuaded
`
`the ION district court that the specification’s references to an “inventive control
`
`system” did not limit the claim scope because they only related to “the preferred
`
`embodiment of the present invention.” Ex. 1126 at 7 (emphasis in original). The
`
`specification corroborates this view, as it twice makes clear that its discussion
`
`pertains to “the preferred embodiment of the present invention.” Ex. 1001 at 3:56-
`
`60, 4:15-19. The newest case cited by WG, Pacing Techs., LLC v. Garmin Int’l,
`
`Inc., 778 F.3d 1021, 1024 (Fed. Cir. 2015), POR at 10, did not change the law that
`
`WG cited to the ION court. It restates that, even under the Phillips standard, “clear
`
`and unmistakable” statements are required to read limitations from the
`
`specification into the claim. Id. WG does not even attempt to identify the express
`
`disclaimer required under the BRI standard, Bigio, 381 F.3d at 1325, and its
`
`arguments must be rejected. The Board’s construction does not require a particular
`
`7
`
`

`
`
`
`prediction methodology, encompasses all of the disclosed embodiments, and is
`
`correct.
`
`B.
`
`“Calculating Desired Changes” in Position Cannot Require
`Calculating Forces or a Particular Method of Calculating Forces.
`
`Limitation (c) of claim 15 recites a “control unit adapted to use the predicted
`
`positions to calculate desired changes in positions of one or more of the [SPDs].”
`
`The specification makes clear that in different embodiments, the controller can
`
`calculate desired changes in force or position. See Ex. 1001 at 6:27-48; Ex. 1117
`
`at 184-86. Obviously, as Dr. Triantafyllou admits, forces and positions are
`
`different things with different units, Ex. 1117 at 186. Because claim 15 is directed
`
`to the latter and not to the former, WG’s argument that claim 15 requires
`
`calculating forces, see POR at 11, would literally rewrite the claim, contrary to
`
`decades of claim construction precedent. Phillips, 415 F.3d at 1314-15; Corning
`
`Glass Works v. Sumitomo Elec. U.S.A., Inc., 868 F.2d 1251, 1257 (Fed. Cir. 1989).
`
`WG also argues that its force calculation “must take into account the
`
`behavior of the array and streamers.” POR at 11-12. Both passages quoted by WG
`
`to support this argument—Ex. 1001 at 4:28-34 and 4:48-51—concern a preferred
`
`embodiment and therefore cannot be used to read limitations into the claim.
`
`Linear Tech. Corp. v. ITC, 566 F.3d 1049, 1058 (Fed. Cir. 2009); Bigio, 381 F.3d
`
`at 1325. The claim language is directly inconsistent with a requirement that the
`
`positions of all of the SPDs must be predicted. And reading in a further limitation
`
`8
`
`

`
`
`
`regarding how the control unit performs the required calculation is especially
`
`inappropriate because claim 15 already says what must be used—the “predicted
`
`positions” generated by the prediction unit of limitation (b). Phillips, 415 F.3d at
`
`1314-15. Dr. Triantafyllou even acknowledged that WG’s proposed construction
`
`finds no basis in the claim language. Ex. 1117 at 234-235.
`
`For purposes of this proceeding, this claim language does not require any
`
`construction beyond rejecting WG’s effort to rewrite it. Limitation (c) specifies
`
`what must be calculated (changes in position) and how (using the predicted
`
`positions). Its broadest reasonable interpretation requires nothing more.
`
`C. The Meaning of “Global Control System” Is the Same Across the
`’607 and ’967 Patents.
`
`Dependent claim 19 recites a “global control system.” Both proposed
`
`constructions of this term are identical to the ones proposed for it in IPR2014-
`
`01475. Compare Pet. at 23-24 with -01475, Pet. at 24-26; POR at 12-17 with -
`
`01475, POR at 9-15. “Because the Board should construe claims consistently
`
`across a family of related patents,” Belden Inc. v. Berk-Tek LLC, 610 F. App’x
`
`997, 1002 (Fed. Cir. 2015) (citing NTP Inc. v. RIM, Ltd., 418 F.3d 1282, 1293
`
`(Fed. Cir. 2005)), the construction of “global control system” adopted by the Board
`
`in -01475 should also be applied here. The Board should adopt PGS’s broader
`
`construction of “global control system” because it is consistent with the patent,
`
`WG’s expert agreed that it is not unreasonable, and WG is judicially estopped from
`
`9
`
`

`
`
`
`challenging it. See e.g., Ex. 1001 at 4:15-17; 4:28-29; 4:48-51; Ex. 1002 ¶¶ 112-
`
`15;
`
` Ex. 1024 at 16-19; IPR2014-00687, Paper 26 at 27; Ex.
`
`1117 at 137, 148, 154-73, 172-173; Ex. 1103 at 122, 125-27, 149-50;
`
`
`
`
`
`II. THE GIKAS ARTICLE (EX. 1006)
`The Gikas article discloses a system for estimating the location of any
`
`element in a seismic array in real-time. Ex. 1002 ¶¶ 131-33. As the Board
`
`recognized in its Institution Decision, the relevance of Gikas to PGS’s obviousness
`
`case is that it teaches limitation (b)’s “prediction unit,” which the POSA would
`
`have combined with the steering system disclosed in the ’636 PCT. Paper 18 at
`
`25-26. WG’s lead arguments about the reference—it “never mentions lateral
`
`control” or “any methods of steering streamers,” POR at 17-18—are inapposite,
`
`because those limitations are disclosed by the ’636 PCT. And as shown below,
`
`WG’s other criticisms of Gikas were disavowed by Dr. Triantafyllou. In the end,
`
`there is no dispute among the experts concerning Gikas’ teaching.
`
`A. Gikas Uses a Dynamic (But Not “Behavior-Predictive”) Model to
`Estimate the Location of Any Point in the Streamer Array.
`
`A Kalman filter, like that disclosed in Gikas, involves two models. Ex. 1006
`
`at 15. The first model uses observations to determine a parameter of interest, such
`
`as position. Id. The second uses that position determination in one time period (or
`
`“epoch”) to generate estimates for it in later time periods. Id.; Ex. 1117 at 264.
`
`10
`
`

`
`
`
`The Gikas Kalman filter is concerned with streamer modeling, and thus its
`
`first (or “primary” or “measurement”) model concerns using the observed positions
`
`(or compass headings) of components of the array to create models of individual
`
`streamers. Ex. 1006 at 16-20. More specifically, Gikas discloses using an n-order
`
`polynomial (preferably fifth or sixth) to make a “highly realistic” estimate of the
`
`position of each streamer. Id. at 14; Ex. 1117 at 270-71. Because Gikas generates
`
`a model of each streamer, the position of any point on any streamer in the array can
`
`be estimated. Ex. 1006 at 16, 24; Ex. 1002 ¶¶ 131-32; Ex. 1117 at 277.
`
`The second (or “dynamic”) model uses a prior position to generate an
`
`updated position estimate for any point in the array. Ex. 1006 at 20. The dynamic
`
`model used in Gikas is relatively straight-forward—it takes a prior position (e.g.,
`
`Xi-1 or Yi-1) and adds to it the product of the velocity (Ẋi-1 or Ẏi-1) and the elapsed
`
`amount of time (dt). Id.; Ex. 1117 at 266-69. Gikas’ use of a prior position,
`
`velocity, and the elapsed time to estimate a present position mirrors the evidence
`
`on which WG relied to persuade the jury that ION’s system met the “predicted
`
`positions” limitation. See Ex. 2087 at 1511-15; Ex. 1128 at 9-10.
`
`The experts agree: the Gikas Kalman filter is a “dynamic” model (i.e., it
`
`accounts for time, see Ex. 1117 at 203) that can use prior positions to estimate the
`
`actual location of any point in the streamer array. Ex. 2052 at 300-01; Ex. 1117 at
`
`265, 277. As Dr. Triantafyllou testified, the Gikas Kalman filter falls squarely
`
`11
`
`

`
`
`
`within the set of “dynamic” (but not “behavior-predictive”) models. Ex. 1117 at
`
`265; Ex. 1105. Given that such dynamic models are disclosed in the ’607 patent’s
`
`specification as preferred embodiments, see Ex. 1001 at 4:28-29, there can be no
`
`question that the Gikas Kalman filter “predicts positions.”
`
`B. Gikas Discloses a Kalman Filter that Operates in Real-Time.
`The experts also agree about how the Gikas Kalman filter would operate in
`
`practice. Dr. Evans explained that such Kalman filters were known to operate in
`
`“real-time.” Ex. 1002 ¶¶ 73-74. And Dr. Triantafyllou agreed that the entire
`
`purpose of the Kalman filter disclosed in Gikas was to provide estimated positions
`
`in the field, not after the survey. Ex. 1117 at 249-50 (discussing Ex. 1006 at 12).
`
`WG has argued that the Gikas Kalman filter “can only be built and tuned
`
`after the entirety of the survey data is collected.” POR at 33. WG cites no expert
`
`testimony to support this contention, see id., nor could it have, given that both Drs.
`
`Evans and Triantafyllou testified that the Gikas Kalman filter generates position
`
`estimates during the survey. Ex. 1002 ¶¶ 73-74; Ex. 1117 at 249-50. Instead of
`
`relying on actual evidence, WG sets forth the following argument regarding Gikas:
`
`“[T]wo stochastic models have to be specified before a Kalman filter can be
`
`implemented . . . Tables 4.10 and 4.20 [entire survey data] . . . [are] used to
`
`develop the stochastic models.” POR at 33 (ellipses and paraphrasing in POR).
`
`WG’s brazen rewriting of Gikas cannot be ignored. The “two stochastic
`
`12
`
`

`
`
`
`models” are not, contrary to WG’s familiar misuse of brackets and ellipses, based
`
`on the “[entire survey data].” They refer to the two covariance matrices that
`
`“describe the quality of the measurements and how well the model describes
`
`reality.” Ex. 1006 at 20. (These matrices are used by the Kalman filter to connect
`
`the primary and secondary models, see Ex. 1006 at 20.) The Gikas article
`
`explicitly states that Tables 4.10 and 4.20, which are used to develop the stochastic
`
`models, “give a full description of the standard deviations used to develop the
`
`stochastic models adopted for these tests.” Id. at 22 (emphasis shows text WG
`
`omitted). These standard deviations are not the “[entire survey data].” They are
`
`error measurements that are based on modeling assumptions and equipment
`
`calibration. See id. at 21-22 and Tables 4.10, 4.20. For that reason, WG’s
`
`argument that Gikas’ Kalman filter cannot be used to estimate positions until after
`
`the survey is plainly incorrect, as its own expert conceded. Ex. 1117 at 249-50.
`
`Likewise, WG’s mischaracterization of Workman cannot be ignored. WG
`
`argues that it is “false” to say that Workman’s discussion of “real time positions”
`
`refers to using a Kalman filter to move old data forward in time. POR at 19, fn.3.
`
`But Workman says precisely that, in its background:
`
`Location sensing devices and methods for determining the positions
`of the seismic sources and seismic streamer cables are also well
`known in the art. [E.g., GPS, acoustic networks.] These devices and
`methods may then be used to determine the real time position of the
`
`13
`
`

`
`
`
`seismic sources and seismic streamer cables by computing a network
`solution to a Kalman filter[.]
`
`Ex. 1004 at 2:10-19. Workman then uses the Kalman filter outputs in repositioning
`
`the streamers in real-time during the survey. Ex. 1004 at 3:45-4:8 & Fig. 3.
`
`WG cites Dr. Cole’s deposition out of context to argue that he “admitted”
`
`that only WG recognized the time lag between positions measured using
`
`techniques such as GPS and acoustics and solving for positions. POR at 7 (citing
`
`Ex. 2045 at 387). The colloquy began several pages earlier when Dr. Cole was
`
`asked whether there was a “delay problem relative to the 5-second delay between
`
`the taking of measurements and the determination of actual streamer positions.”
`
`Ex. 2045 at 380-81. Dr. Cole then summarized how the Workman system
`
`functioned, id. at 381-85, how the Kalman filter (mistranscribed as “common
`
`filter”) operated in real time within the 10-second window between shots, id. at
`
`385-86, and that the POSA would have been able to implement a system that was
`
`“quick enough to do it between shots,” id. at 386. With that context, in the
`
`testimony WG mischaracterizes, Dr. Cole explained that he did not believe the
`
`POSA would have considered the delay between measurement and determination
`
`to be a problem because the prior art already possessed a solution in the form of
`
`the Kalman filter (citing Workman and the Gikas thesis). See id. at 386-87.
`
`III. THE CHALLENGED CLAIMS ARE OBVIOUS
`There is little dispute that under the Board’s correct constructions of the
`
`14
`
`

`
`
`
`claims, they are obvious. WG did not elicit any expert testimony to defend claim
`
`15 under the Board’s constructions, despite having every opportunity to do so. The
`
`reason for this is clear—not even its long-serving expert was willing to so opine.
`
`In trying to defend these claims at his deposition, Dr. Triantafyllou clung to his
`
`“behavior-predictive model” requirement like a life raft, infusing it as a new
`
`limitation into every disputed claim of all three disputed patents, even where there
`
`was no linguistic hook in the claims to support it. Ex. 1118 at 377-80, 493-94,
`
`577-80. But whether a claim directed to Dr. Triantafyllou’s vision of a behavior-
`
`predictive model, taking into account all of the forces on all of the streamers in the
`
`array, would have been novel and non-obvious—or for that matter enabled given
`
`the five-word description in the specification and the myriad algorithms Dr.
`
`Triantafyllou contends are required, Ex. 1103 at 165-70; Ex. 1117 at 320-23—is
`
`entirely irrelevant. That is not what WG described or claimed. It is not what WG
`
`asserted in its infringement case against ION. And it is not what WG has told the
`
`EPO is its invention.
`
`A. Modifying the ’636 PCT’s System To Use Gikas’ Kalman Filter
`Would Have Been Obvious, Making Claim 15 Obvious.
`
`Construed correctly, claim 15 recites an array of towed streamers comprising
`
`three components: (a) two or more SPDs on or inline with each of two or more
`
`seismic streamers; (b) a prediction unit that can estimate the actual locations of at
`
`least two of the SPDs; and (c) a control unit that uses those positions to calculate
`
`15
`
`

`
`
`
`desired changes in position of at least one SPD. See supra, § I.
`
`The system disclosed in the ’636 PCT teaches every limitation except (b),
`
`the prediction unit. Pet. at 28-29, 33-35; Ex. 1002 ¶¶ 126-29, 134-40. In the ’636
`
`PCT, the SPD positions were determined using the least-squares method. Ex. 1002
`
`¶ 134. As explained in Gikas, the least-squares method solves for positions at a
`
`given time (“epoch”), but does not incorporate information from prior epochs. Ex.
`
`1006 at 12. Accordingly, the ’636 PCT’s system for determining SPD locations is
`
`arguably not capable of “predicting positions,” i.e., estimating the locations of the
`
`SPDs, because it is determining where they actually were a few seconds earlier.
`
`Dr. Evans testified, however, that the POSA would have had a reason to modify
`
`the ’636 PCT’s system to replace its least-squares-based system with Gikas’
`
`Kalman filter, which as explained above is a “prediction unit” capable of
`
`predicting the location of any point in the streamer array, including the positions of
`
`the SPDs that are on or inline with a streamer. Ex. 1002 ¶¶ 130-34, 138-40; see
`
`supra § II-A. Dr. Evans’ testimony stands unrefuted; Dr. Triantafyllou did not
`
`dispute his analysis on this point.2 Lacking expert testimony of its own, WG
`
`derides Dr. Evans’ analysis as “conclusory,” see POR at 31, but fails to engage or
`
`2 His only response was that the POSA would not have been motivated to modify
`
`Gikas’ Kalman filter to implement “behavior-predictive model-based control.” See
`
`Ex. 2075 ¶ 166. Under the correct claim construction, this testimony is irrelevant.
`
`16
`
`

`
`
`
`refute that the POSA would have accepted Gikas’ teaching and thorough
`
`explanation as to why its Kalman filter was an improvement over least-squares for
`
`determining the positions of any point in a streamer array. See Ex. 1002 ¶ 134.
`
`The motivation for the combination is explicit in the Gikas article, where Dr. Gikas
`
`explained no fewer than seven benefits to determining positions using his Kalman
`
`filter rather than the prior art least-squares method. See id.; Ex. 1006 at 12-13.
`
`WG and its expert also fail entirely to refute the testimony of PGS’s experts,
`
`Drs. Evans and Cole, that implementing the Gikas Kalman filter in the ’636 PCT’s
`
`system would have been within the abilities of the POSA. Dr. Evans explained
`
`that using positions predicted