`571-272-7822
`
`Paper 17
`Date: February 9, 2022
`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`APPLIED MATERIALS, INC.,
`Petitioner,
`v.
`OCEAN SEMICONDUCTOR LLC,
`Patent Owner.
`
`IPR2021-01342
`Patent 6,968,248 B1
`
`
`
`
`
`
`
`
`
`Before MIRIAM L. QUINN, JOHN D. HAMANN, and DAVID COTTA,
`Administrative Patent Judges.
`QUINN, Administrative Patent Judge.
`
`DECISION
`Granting Institution of Inter Partes Review
`35 U.S.C. § 314, 37 C.F.R. § 42.4
`
`
`
`
`
`
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`
`I.
`INTRODUCTION
`Applied Materials, Inc. (“Petitioner”) filed a Petition (Paper 1,
`“Petition” or “Pet.”) requesting an inter partes review of claims 1–22 (“the
`challenged claims”) of U.S. Patent No. 6,968,248 B1 (Ex. 1001, “the ’248
`patent”) pursuant to 35 U.S.C. §§ 311–319. Ocean Semiconductor LLC
`(“Patent Owner”) filed a Preliminary Response. Paper 10 (“Preliminary
`Response” or “Prelim. Resp.”). With our authorization, Petitioner filed a
`Reply to Patent Owner’s Preliminary Response (Paper 13, “Reply”), and
`Patent Owner filed a Sur-Reply in Support of Patent Owner’s Preliminary
`Response (Paper 14, “Sur-reply”).
`The standard for institution is set forth in 35 U.S.C. § 314, which
`provides that an inter partes review may not be instituted unless the
`information presented in the Petition and the Preliminary Response shows
`that “there is a reasonable likelihood that the petitioner would prevail with
`respect to at least 1 of the claims challenged in the petition.” 35 U.S.C.
`§ 314 (2018); see also 37 C.F.R § 42.4(a) (“The Board institutes the trial on
`behalf of the Director.”). Upon consideration of the parties’ contentions and
`the evidence of record, we conclude that Petitioner has established a
`reasonable likelihood of prevailing in demonstrating the unpatentability of at
`least one challenged claim of the ’248 patent. Accordingly, we grant
`Petitioner’s request and institute an inter partes review of the challenged
`claims.
`
`A. Related Matters
`The parties indicate that the ’248 patent has been asserted in the
`following proceedings: Ocean Semiconductor LLC v. Analog Devices, No.
`1:20-cv-12310 (D. Mass); Ocean Semiconductor LLC v. Infineon, No. 1:20-
`cv-12311 (D. Mass.); Ocean Semiconductor LLC v. Huawei, No. 4:20-cv-
`
`2
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`911 (E.D. Tex.); Ocean Semiconductor LLC v. MediaTek, No. 6:20-cv-1210
`(W.D. Tex.); Ocean Semiconductor LLC V. NVIDIA, No. 6:20-cv-1211
`(W.D. Tex.); Ocean Semiconductor LLC v. NXP, No. 6:20-cv-1212 (W.D.
`Tex.); Ocean Semiconductor LLC v. Renesas, No. 6:20-cv-1213 (W.D.
`Tex.); Ocean Semiconductor LLC v. Silicon Labs, No. 6:20-cv-1214 (W.D.
`Tex.); Ocean Semiconductor LLC v. ST Micro, No. 6:20-cv-1215 (W.D.
`Tex.); and Ocean Semiconductor LLC v. Western Digital, No. 6:20-cv-1216
`(W.D. Tex.). Pet. 1–2; Paper 5, 2.
`B. The ’248 Patent
`The ’248 patent relates to “scheduling in an automated manufacturing
`environment.” Ex. 1001, 1:20–21. The ’248 patent describes the
`manufacture of integrated circuits for modern semiconductor devices
`containing numerous structures or features, typically the size of a few
`micrometers. Id. at 1:38–41. The ’248 patent further describes that the
`fabrication of integrated circuits generally involves processing a number of
`wafers through a series of fabrication tools, where layers of material are
`added to, removed from, and/or treated on a semiconducting substrate. Id. at
`1:41–45. According to the ’248 patent, controlling a semiconductor factory
`(“fab”) that fabricates such integrated circuits is a challenging task, where
`the fab is a complex environment where numerous parts (typically 40,000
`wafers or more) and numerous part types (typically 100 part types or more)
`are simultaneously being manufactured. Id. at 1:65–2:3. As each wafer
`moves through the fab, it may undergo more than 300 processing steps,
`many of which use the same machines, where a large factory may contain
`approximately 500 computer-controlled machines to perform this wafer
`processing. Id. at 2:3–8. As described in the ’248 patent, routing,
`scheduling, and tracking material through the fab is a difficult and
`
`3
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`complicated task, even with the assistance of a computerized factory control
`system. Id. at 2:8–11.
`Figure 3 illustrates an implementation of reactive scheduling of
`activities of a process flow for a semiconductor fabrication facility and is
`reproduced below.
`
`
`
`Figure 3 shows a portion of process flow 100 from a semiconductor
`fabrication facility, and the manner in which it schedules appointments for
`the consumption of resources. Id. at 4:28–32. Process flow 100 includes
`stations 105, each station 105 including computing device 110
`communicating with process tool 115. Id. at 5:17–19. Process tools 115 are
`processing lots 130 of wafers 135 that will eventually become integrated
`circuit devices, where process tool 115 may be a fabrication tool used to
`fabricate some portion of wafers 135. Id. at 5:24–26, 6:43–45.
`Each computing device 110 includes software agent 265, where
`software agents 265, collectively, are responsible for efficiently scheduling
`
`4
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`and controlling lots 130 of wafers 135 through the fabrication process. Id. at
`6:24–26, 47–50. Collectively, software agents 265 reactively and
`proactively schedule activities for each lot 130 for operations on a specific
`qualified process tool 115. Id. at 6:63–7:3. More specifically, the software
`agents (or scheduling agents) 265 include: Lot Scheduling Agent (“LSA”)
`305 that schedules activities on behalf of lots 130 of wafers 135; Machine
`Scheduling Agent (“MSA”) 310 that schedules activities on behalf of
`process tools 115; PM Scheduling Agent (“PMSA”) 315 that schedules
`activities on behalf of preventative maintenance (“PMs”) and equipment
`qualification (“Quals”) (not shown in Figure 3); and Resource Scheduling
`Agent (“RSA”) that schedules activities on behalf of resources (not shown in
`Figure 3). Id. at 7:20–30. Some of these activities are scheduled reactively
`(i.e., in response to events occurring in process flow 100). Id. at 7:36–37.
`For example, the ’248 patent describes the process as detecting an
`occurrence of a predetermined event in the process flow 100; notifying a
`subscribing software scheduling agent (e.g., LSA 305, MSA 310, PMAS
`315, or RSA 320) of the occurrence; and reactively scheduling an action
`responsive to the detection of the predetermined event. Id. at 7:38–46.
`C. Illustrative Claims
`Of the challenged claims, claims 1 and 14 are independent. Each of
`challenged claims 2–13 and 15–22 depends from claim 1 or 14.
`Claim 1 is illustrative:
`scheduling
`for
`1. A method
`manufacturing environment, comprising:
`automatically detecting an occurrence of a predetermined
`event in an integrated, automated process flow;
`automatically notifying a software scheduling agent of the
`occurrence; and
`
`in an automated
`
`5
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`the software
`reactively scheduling an action from
`scheduling agent
`responsive
`to
`the detection of
`the
`predetermined event.
`Ex. 1001, 30:40–48.
`D. Asserted Prior Art and Ground of Unpatentability
`The asserted ground in this proceeding involves the following prior art
`references:
`a) Schulze: US 2002/0116083, published August 22, 2002, filed as
`Exhibit 1007; and
`b) Gupta: US 4,888,692, issued December 19, 1989, filed as Exhibit
`
`1008.
`
`Petitioner asserts the following ground of unpatentability (Pet. 16):
`Claim(s) Challenged
`35 U.S.C. §
`Reference(s)/Basis
`1–22
`103
`Schulze, Gupta
`Petitioner also relies on a Declaration of Dr. Stanley Shanfield, filed
`as Exhibit 1003 (“Shanfield Declaration” or “Shanfield Decl.”).
`II. ANALYSIS
`A. Level of Ordinary Skill in the Art
`Petitioner contends that a person having ordinary skill in the art
`“would have at least a B.S. in computer science, mechanical engineering,
`electrical engineering, or a related field, and three years of experience
`working with automated manufacturing processes.” Pet. 17. For purposes
`of its Preliminary Response, Patent Owner does not dispute this contention.
`Prelim. Resp. 36. At this juncture, we do not find it necessary to define the
`level of skill with specificity save to note that the level of ordinary skill is
`evidenced by the prior art of record. See Okajima v. Bourdeau, 261 F.3d
`1350, 1355 (Fed. Cir. 2001) (stating that the absence of specific findings on
`the level of skill in the art does not give rise to reversible error where the
`
`6
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`prior art itself reflects an appropriate level and a need for testimony is not
`shown).
`
`B. Claim Construction
`In inter partes review proceedings based on petitions filed on or after
`November 13, 2018, such as this one, we construe claims using the same
`claim construction standard that would be used in a civil action under
`35 U.S.C. § 282(b), as articulated in Phillips v. AWH Corp., 415 F.3d 1303
`(Fed. Cir. 2005) (en banc), and its progeny. See 37 C.F.R. § 42.100(b).
`Petitioner submits that no claim term needs to be construed, as the
`prior art relied on in the Petition discloses the subject matter of the
`challenged claims under any reasonable construction, including their plain
`meaning. Pet. 16. Patent Owner contends that claim construction is not
`necessary for the Board to make a decision on institution. Prelim. Resp. 36.
`We need not determine preliminarily the construction for any claim term as
`our decision does not rely on any.
`C. Obviousness over Schulze and Gupta
`We now analyze the ground asserted in the Petition—that claims 1–22
`would have been obvious over Schulze and Gupta. Pet. 33–67.
`1. Overview of Schulze (Ex. 1007)
`Schulze is related to “systems and methods for monitoring and
`assessing the performance and operation of . . . semiconductor fabrication
`facilities.” Ex. 1007 ¶ 3. Schulze describes a need to identify which
`fabrication tools of large semiconductor fabrication facilities are idle in
`order to reduce the idle time of the fabrication tools and maximize
`production time, yield and profitability. Id. ¶¶ 6, 7.
`Figure 1 illustrates a semiconductor fabrication system incorporating
`an automated monitoring and assessment system and is reproduced below.
`
`7
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`
`
`
`Figure 1 is a top-level diagram illustrating semiconductor fabrication
`system 100 incorporating an automated monitoring and assessment system.
`Id. ¶ 40. As illustrated in Figure 1, manufacturing execution system 102 is
`connected to system bus 105, along with a plurality of semiconductor
`fabrication tools 115. Id. Also connected to system bus 105 is automated
`monitoring and assessment system 107 and bus controller 109. Id. Bus
`controller 109 controls the routing of information over system bus 106, and
`automated monitoring and assessment system 107 subscribes to the
`information needed for performing the monitoring and assessment functions.
`Id. ¶ 41. Semiconductor fabrication tools 115 transmit or publish messages
`over system bus 105 to manufacturing execution system 102. Id. ¶ 42.
`
`8
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`Automated monitoring and assessment system 107 receives information
`from the transmitted or published messages, and uses that information to
`track the operation states of the semiconductor fabrication tools 115. Id.
`2. Overview of Gupta (Ex. 1008)
`Gupta “relates to automated scheduling and planning systems.”
`Ex. 1008, 1:9–10. More specifically, Gupta describes a system for
`scheduling the operation of interrelated machines (such as a manufacturing
`facility for integrated circuits), which perform a process flow. Id. at [57],
`3:50–51. A global definition of the system is made once, and each machine
`has an individual profile describing its local interaction with the system. Id.
`at [57]. However, local scheduling decisions for each machine are made
`based on that machine’s individual profile and the state of the manufacturing
`facility at the time a decision is needed. Id.
`Gupta states that the “real-time portion of the scheduling system
`depends on local optimization to function efficiently.” Id. at 13:4344.
`Thus, Gupta explains, instead of recalculating the complete global state for
`the system, each time a decision must be made, only the relevant local state
`is recalculated. Id. at 13:4448. Such a process reduces processor load. Id.
`Each machine in the manufacturing facility has several data structures
`that determine the machine behavior. Id. at 13:4952. “Decision-making is
`event driven, and a determination of what comes next for each machine is
`made whenever certain events take place.” Id. at 13:5456. Gupta identifies
`events that drive the decision making process: machine loads and unloads, a
`machine going off-line and coming on-line. Id. at 13:5658. According to
`Gupta, whenever one of these events occurs, “the scheduling system must
`calculate what that machine will do next.” Id. at 13:5960. Those actions,
`
`9
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`Gupta explain, may be to load a lot immediately and to take the lot from an
`input queue. Id. at 13:6164. Other actions may involve requiring a
`machine to wait for a full load, or to proceed with a partial load. Id. at
`13:6466.
`Gupta’s scheduler, therefore, makes decisions when machines are due
`to load, or when they unload. Id. at 14:1921. “Since the schedule knows in
`advance when its computational resources will be in demand, it is in a
`position to look ahead and predict when its resources will be inadequate to
`fully compute each required decision.” Id. at 14:2126. “If a heavy demand
`on computational resources will be required at some time in the future, the
`scheduling system will need to begin making decisions ahead of time.” Id.
`at 14:3033.
`
`3. Reasonable Likelihood Determination
`After considering Petitioner’s contentions and Patent Owner’s
`arguments in opposition, and as we discuss below, we are persuaded that
`Petitioner has demonstrated a reasonable likelihood of prevailing on
`showing that claims 1–22 would have been obvious over Schulze and Gupta.
`a) Independent Claim 1
`Petitioner contends that Schulze describes an automated
`manufacturing environment with a non-intrusive, reliable, and
`comprehensive system or method for monitoring, assessing, and reporting
`the operation and performance of a semiconductor manufacturing facility (or
`“fab”) by connecting an automated monitoring and assessment system to a
`manufacturing execution system (MES) and fabrication tools via a system
`bus. Pet. 34 (citing Ex. 1007 ¶¶ 14, 40; Shanfield Decl. ¶ 111); Pet. 41
`(citing Ex. 1007 ¶¶ 3940); Shanfield Decl. ¶ 131. By disclosing that the
`
`10
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`MES is connected to the system bus, along with a plurality of semiconductor
`fabrication tools, and disclosing that the MES monitors the connected
`fabrication tools activity by detecting changes on the machine operation,
`such as unscheduled downtime, Petitioner contends that Schulze teaches the
`limitation of claim 1 that requires “automatically detecting an occurrence of
`a predetermined event in an integrated, automated process flow.” Pet. 42
`(citing Ex. 1007 ¶¶ 40, 55, 214; Shanfield Decl. ¶ 133). This assessment by
`Petitioner has not been challenged by Patent Owner at this juncture. We
`find that these arguments and the evidence presented for this “automatically
`detecting” limitation meet the reasonable likelihood threshold of institution.
`For the remainder of the limitations of claim 1, Petitioner relies on
`Gupta’s teachings. And Patent Owner contests Gupta’s role in Petitioner’s
`case. We discuss each of the remaining limitations of claim 1 separately
`below.
`(1) “automatically notifying a software scheduling agent of the occurrence”
`(“automatically notifying” limitation)
`Petitioner argues that Gupta discloses a scheduler implemented in
`software that constitutes the claimed “scheduling agent.” Pet. 4243 (citing
`Ex. 1008, 30:316; Shanfield Decl. ¶¶ 65, 7281, 134). According to
`Petitioner, Gupta’s scheduler makes decisions for a given machine in the
`process flow by applying the “local prediction” concept. Id. at 43 (citing
`Ex. 1008, 1:3448, 3:5766, 14:826, 16:817, 17:713, 20:58, 20:514,
`25:2241, 25:5255; Shanfield Decl. ¶¶ 134). Therefore, Petitioner argues,
`Gupta’s decision-making is on a real-time basis, requiring the scheduler to
`be automatically notified of the event detected at the fabrication facility. Id.
`(citing Ex. 1008, 2:5152; Shanfield Decl. ¶ 135). Patent Owner does not
`contest these assertions.
`
`11
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`Petitioner’s arguments and evidence at this juncture reasonably
`convey how Gupta in combination with Schulze teaches this “automatically
`notifying” limitation of claim 1. For instance, the Petition explains that
`implementing Gupta’s scheduler in Schulze’s automated monitoring and
`assessment system 107, which, in combination with Schulze’s MES102 and
`system bus 105, would automatically receive the events detected about the
`various equipment 115. Pet. 4345. We find that these arguments, and the
`evidence presented for this “automatically notifying” limitation, meet the
`reasonable likelihood threshold of institution.
`(2) “reactively scheduling an action from the software scheduling agent
`responsive to the detection of the predetermined event” (“reactively
`scheduling” limitation)
`Petitioner asserts that Gupta’s scheduler, implemented as stated above
`in Schulze’s automated monitoring and assessment system, performs a
`“local optimization” to reactively schedule an action to be performed by a
`machine. Pet. 45–46 (citing Ex. 1008, 13:5660, 15:4954, 19:544,
`22:5763, 23:3451; Shanfield Decl. ¶¶ 139140). For instance, Petitioner
`points out Gupta’s disclosure that Gupta receives a notification that a
`machine just unloaded 2 out of 4 lots in its queue, and that based on that
`event, Gupta’s scheduler performs local optimization to decide “whether to
`load those 2 lots now, or to wait some short period of time until 1 or 2 more
`lots arrive so that a larger load can be processed.” Pet. 46 (citing Ex. 1008,
`15:4954; Shanfield Decl. ¶ 140). Petitioner proffers additional examples of
`actions scheduled by Gupta’s scheduler in response to notifications from a
`machine. Id. (explaining how Gupta’s scheduler revises the upstream
`processing schedule in response to a machine coming online after a repair
`
`12
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`and Gupta’s scheduler may delay the upstream machine upon notification of
`a machine’s breakage).
`Patent Owner presents several arguments challenging Petitioner’s
`reliance on Gupta for this “reactively scheduling” limitation. First, Patent
`Owner argues that Gupta is “predictive,” not “reactive” as the claim
`requires. Prelim. Resp. 4347. Second, Patent Owner argues that Gupta
`teaches away from the asserted combination with Schulze. Id. at 3743.
`Third, Patent Owner argues that Petitioner has relied on impermissible
`hindsight because the record lacks support for the asserted combination. Id.
`at 4243. We address each argument in turn.
`(a) Patent Owner’s Predictive Versus Reactive Argument
`Gupta states that “[w]hen a machine must undertake local
`optimization, it runs a local simulation to determine what the future will
`bring.” Ex. 1008, 20:1214. According to Patent Owner, Gupta uses
`prediction, even in cases where the existing scheduling for a given machine
`no longer makes sense due to breakdowns. Prelim. Resp. 4344. In short,
`Patent Owner argues, Gupta “requires an intermediate predictive step,”
`“before any scheduling is completed.” Id. at 44 (emphasis omitted).
`Because Gupta performs a prediction or recalculation in its decision-making
`process, the argument goes, Gupta is “predictive” and not reactive as the
`claim requires. Id. at 4445. At this juncture, we do not agree with Patent
`Owner’s contention.
`As Petitioner asserts, Gupta’s local optimization is an event-driven
`decision-making process. Ex. 1008, 13:5460. That decision-making
`process includes making predictions in advance, when computational
`resources are expected to be scarce. Id. at 14:2125, 16:47. And it also
`
`13
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`includes making a local prediction or simulation in real time, such as when
`the machine receives a demand signal. Id. at 19:4559. That Gupta uses the
`word “prediction” to refer to the calculations it performs in response to an
`event does not appear to contradict or be the opposite of “reactively
`scheduling” as Patent Owner argues. Gupta’s “prediction” refers to
`calculations that consider the impact of future processing steps that may
`affect the local machine. See id. at 20:514. The local prediction process
`“consists of asking the upstream processes what they will be doing in that
`time frame,[ a few tens of time steps away,] and applying the decision
`making process to the results.” Id. at 20:1520. In case of events that are
`unpredicted, such as machine breakdowns, the calculations must be repeated
`upon receiving a demand signal. Id. at 20:2227. Gupta’s local prediction
`thus appears to refer to obtaining and analyzing the data needed to calculate
`which action to take. The word “prediction” in Gupta alludes to the nature
`of the data analysis, being future-focused, such as tabulating what each
`machine is in the process of handling and the expected queue. See, e.g., id.
`at 22:326, Table 5 (describing the local prediction process for multi-lot
`machine optimization, which applies lot function minimization to the
`predicted lot and processing time for the machine).
`But when Gupta refers to determining what actions to take, it appears
`to do so “reactively,” by responding to the events occurring in the machine.
`See Ex. 1001, 7:3537 (explaining that activities scheduled “reactively” are
`scheduled “in response to events occurring in, e.g., the process flow 100, in
`accordance with the present invention”). For example, Gupta explains that
`when a machine M1 is down, process P37 sends a negative request signal to
`P36 using a routine that places the negative request signal in the data
`
`14
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`structures for the upstream processes. Ex. 1008, 24:616; see also Pet. 46
`(citing Ex. 1008, 23:3451, describing the negative request signal sent
`upstream to delay the original schedule for a certain amount of time). Thus,
`despite using prediction analysis to determine the processing time and
`queues for each machine, the actions Gupta takes, such as executing a
`routine to distribute negative request signals upstream, are reactive as they
`are taken in response to the event occurring in the machines. Therefore, at
`this juncture, we are not persuaded that Gupta’s local optimization or
`“prediction” calculations contradict the nature of its actions being “reactive”
`as required by the claim.
`Patent Owner further argues that the “cadence of the scheduling
`process” of Gupta is slow and that the computational intervals of Gupta are
`too long for Gupta to be considered reactive. Prelim. Resp. 4647. For
`instance, Patent Owner posits that Gupta is unlike the ’248 patent because
`the change of Gupta’s schedule is preceded by a predictive step and that
`lengthy interval makes Gupta unable to react to factory changes. Id. at 47.
`We do not agree with Patent Owner’s contention.
`First, even if Gupta’s action occurs after it performs the prediction
`calculation, the action is nonetheless reactive, as it is performed in response
`to the received notification. As described above, when a machine M1 is
`down, the routine to distribute a negative request signal upstream occurs in
`response to the machine down event, which is what the claim requires. The
`intervening prediction analysis, even if there were one in the machine down
`situation, does not negate the reactiveness of Gupta’s decision to send a
`negative request signal upstream, according to the claim. And we see no
`
`15
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`claim requirement that limits “reactively scheduling an action” to an action
`that occurs without any intervening calculations.
`Second, we understand Patent Owner’s argument to focus on
`“immediacy” based on embodiments in the ’248 patent. See Prelim. Resp.
`4647 (citing Ex. 1001, Table 3, 13:5358). Without more, however, we do
`not read embodiments as claim limitations. See Phillips v. AWH Corp., 415
`F.3d 1303, 1323 (Fed. Cir. 2005) (warning against confining the claim to
`specific embodiments of the invention). Therefore, we are not persuaded by
`Patent Owner’s arguments that the ’248 patent includes two embodiments
`that use the words “as soon as possible” and “immediately,” yet Gupta’s six-
`minute calculation time is too long in comparison. Here again, we see no
`claim language that requires immediacy between the step of “automatically
`notifying” and “reactively scheduling an action.”
`In sum, Patent Owner’s arguments that Gupta is “predictive” rather
`than “reactive” in terms of the claim limitation “reactively scheduling an
`action” are not persuasive at this juncture.
`(b) Patent Owner’s Argument that Gupta Teaches Away
`Petitioner provides a lengthy discussion on the combinability of
`teachings of Schulze and Gupta. Pet. 3541. We point out for instance
`Petitioner’s argument that a person of ordinary skill in the art “would have
`appreciated that the distributed and localized nature of the software
`scheduler taught in Gupta would reduce the computational complexity
`involved in each scheduling decision.” Id. at 36 (citing Shanfield Decl.
`¶ 118; Ex. 1021, 207; Ex. 1013, 566). Additionally, Petitioner argues that
`Gupta provides the advantage of avoiding the need to reconfigure the
`scheduling strategy for the entire process flow upon adding a new type of
`machine to the process flow because Gupta’s local scheduler can be adjusted
`
`16
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`for the new machine. Id. at 3637 (citing Shanfield Decl. ¶ 119; Ex. 1008,
`14:6061, 30:3234; Ex. 1024, 46).
`In its opposition, Patent Owner contrasts Schulze’s “all tools”
`monitoring with Gupta’s individual machine focus to argue that Gupta
`teaches away from implementing the local scheduler with Schulze’s
`teachings. Prelim. Resp. 3640. For instance, Patent Owner posits that
`Gupta avoids taking actions that would involve the entire factory because
`Gupta teaches “local optimization.” Id. at 3940. Patent Owner further
`argues that Gupta “counsels against attempts to scale up its scheduling
`method to the entire fab,” because Gupta seeks to conserve processing
`power. Id. at 4142. Specifically, Patent Owner focuses on Gupta’s
`disclosure that [d]ecisions which consider many factors, such as those made
`for the entire facility at once, tend to require prohibitive computational
`resources,” and, therefore, a person of ordinary skill in the art would have
`been “actively discouraged” from adding “scheduling to Schulze’s fab-wide
`MES.” Id. at 41 (emphasis omitted).
`We are not persuaded by Patent Owner’s argument. Gupta describes
`“global planning” and “local optimization” for the entire fab. Ex. 1008,
`8:469:21, 13:4114:26. Thus, arguing that Gupta’s scheduler would not be
`focused on an entire fab is belied by Gupta’s disclosure. Further, the
`portions of Gupta that Patent Owner relies on do not support what Patent
`Owner asserts—that Gupta’s scheduler is expressly designed to not work for
`the entire factory. We understand Gupta to teach a system design that makes
`a large number of simple computational decisions for each machine of the
`entire factory, rather than a computation heavy design that requires
`recalculating the global parameters for the entire factory each time a
`
`17
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`decision is made. Id. at 13:6714:11. This is different from how Patent
`Owner characterizes Gupta—i.e., as “implicitly counsel[ing] against
`attempts to scale up its scheduling method to the entire fab.” Prelim. Resp.
`41 (see also Prelim. Resp. 39, arguing unpersuasively that Gupta expressly
`teaches that its local optimization should not be applied to an entire factory
`and likely would not work if it were so applied).
`Furthermore, emphasizing Gupta’s local optimization as limiting the
`number of tools or number of actions is also unpersuasive. First, Gupta
`describes that the “range of actions which can be taken is fairly limited” in
`the context of explaining that the machine actions are rather simple: load a
`lot from its queue immediately, wait for a full load, or proceed with a partial
`load. Ex. 1008, 13:6166. Gupta is not saying that its scheduler is
`somehow limited in actions—rather, the machine’s actions are simple, thus
`the number of options for the scheduler to decide are simple. Second,
`Gupta’s scheduler makes a decision for all the machines in the system by
`local prediction, regardless of the computational resources available. For
`instance, if the scheduler resources are tight, “such as a very large facility
`using a small computer system for scheduling and planning,” the more
`critical machines will have first call on the resources and the other machines
`employ simpler decision strategies or a default strategy, such as round-robin
`mode. Id. at 14:4359. But if the scheduler knows how long it will take
`“for each machine,” it will need to begin making decisions ahead of time,
`such as six minutes in advance, so that when the machines unload, the
`results of the prediction are available to the machine. Id. at 14:2742. Thus
`Gupta explains that its scheduler takes into account each machine and
`performs the expected decisions for each machine, regardless of the
`
`18
`
`IPR2022-00681
`PATENT OWNER, EX 2001
`
`
`
`IPR2021-01342
`Patent 6,968,248 B1
`computational strain. Therefore, the argument that Gupta teaches away from
`applying its scheduler to an entire facility, such as that of Schulze, because
`of its local optimization scheme, is untenable on the present record.
`Consequently, we are not persuaded that Gupta teaches away from
`combining its teachings with those of Schulze as presented in the Petition.
`(c) Patent Owner’s Hindsight Allegation
`In connection with the reasons to combine, Patent Owner alleges that
`Petitioner proffers conclusory arguments with insufficient factual support,
`and, therefore, Petitioner’s theory relies on improper hindsight. Prelim.
`Resp. 4243. We do not agree. Patent Owner takes issue with one of
`several reasons to combine proffered by Petitioner in which Dr. Shanfield
`states that a person of ordinary skill in the art would have recognized that the
`distributed and localized approach of Gupta’s scheduler would reduce the
`computational resources and time needed. Id. at 42 (citing Shanfield Decl.
`¶ 118). According to Patent Owner, Dr. Shanfield does not explain the
`connection between Gupta and a distributed and localized approach, and the
`missing link is found in secondary references, one of which is dated after the
`’248 patent’s priority date. Id. Even if Patent Owner’s argument had merit,
`which does not at this juncture appear to be the case given that Dr.
`Shanfield’s testimony is unrebutted and appears reasonable on its face,
`Petitioner proffers several other explanations bearing on the analysis of
`reasons to combine. This additional explanation provides rational
`underpinning sufficient to meet the requirements of KSR Int’l v. Teleflex,
`550 U.S. 398, 418 (2007). For instance, as stated earlier, one