DCT

1:26-cv-00598

Carnegie Mellon University v. Uber Tech Inc

Log In
Key Events
Complaint
complaint Intelligence

I. Executive Summary and Procedural Information

  • Parties & Counsel:
  • Case Identification: Carnegie Mellon University v. Uber Technologies, Inc., 1:26-cv-00598, D. Del., 05/26/2026
  • Venue Allegations: Venue is alleged to be proper in the District of Delaware because Defendants are Delaware corporations, reside in the District, have transacted business in the District, and have committed acts of infringement there.
  • Core Dispute: Plaintiff alleges that Defendant's dynamic pricing and logistics platform, which utilizes an "H3 hexagon and hexclustering system," infringes four patents related to methods for discovering and analyzing dynamic, data-driven neighborhood clusters.
  • Technical Context: The technology concerns a shift from using static, predefined geographic boundaries (like zip codes or municipal neighborhoods) to using dynamic, activity-based data (like user "check-ins") to define and understand urban areas.
  • Key Procedural History: The complaint alleges the patented technology stems from a 2012 academic paper by the inventors, which received industry accolades. It also alleges that an engineer from Uber contacted the inventors in 2012 to express interest in the project, years before the first patent-in-suit issued.

Case Timeline

Date Event
2012-04-17 Uber engineer allegedly contacted inventors about the "Livehoods" project.
2012-08-30 Earliest Priority Date for '887, '672, '349, and '082 Patents.
2012-XX-XX Uber introduced surge pricing.
2014-XX-XX Uber introduced "neighborhood surge" or "geofence surge."
2017-12-19 U.S. Patent No. 9,846,887 Issued.
2020-XX-XX Uber introduced its "Super App" view.
2020-07-14 U.S. Patent No. 10,713,672 Issued.
2022-01-11 U.S. Patent No. 11,222,349 Issued.
2024-03-19 U.S. Patent No. 11,935,082 Issued.
2026-05-26 Complaint Filed.

II. Technology and Patent(s)-in-Suit Analysis

U.S. Patent No. 9,846,887

  • Patent Identification: U.S. Patent No. 9,846,887 ("Discovering Neighborhood Clusters and Uses Therefor"), issued December 19, 2017 Compl. ¶34

The Invention Explained

  • Problem Addressed: Traditional online maps and location-based services rely on static, geographically-defined organizational units like neighborhoods, which fail to capture the "changing dynamics of a city" and its real-world activity patterns Compl. ¶22 '887 Patent, col. 1:36-40
  • The Patented Solution: The invention proposes systems and methods to discover dynamic, data-driven neighborhood clusters (termed "Livehoods") by analyzing "venue check-in data" Compl. ¶24 These clusters are defined not just by geographic proximity but also by "social similarity," reflecting how people actually move between and interact with different venues, thus revealing the "current collective activity patterns of people in a city" Compl. ¶¶25, 39 Figure 1 of the complaint illustrates how these data-driven "Livehoods" differ from static municipal borders Compl. p. 5 The patent specification details how a "cluster discovery module" processes this data to identify the clusters '887 Patent, col. 4:45-55 '887 Patent, fig. 1
  • Technical Importance: This approach provided an objective, data-driven method to define and analyze urban areas based on actual human behavior, moving beyond subjective or outdated static boundaries Compl. ¶40

Key Claims at a Glance

  • The complaint asserts independent method claim 23 Compl. ¶214
  • Essential elements of Claim 23 include:
    • Storing, in a computer database system, derived venue check-in data based on time-stamped location data.
    • Generating a "check-in intensity vector" for each venue based on the check-in data.
    • Generating elements of a "pairwise venue similarity matrix" for the venues, with a similarity score based on the similarity between the check-in intensity vectors.
    • Determining boundaries for two or more geographic clusters of venues based on the similarity matrix.
    • Transmitting the determined geographic clusters to an analytics server system.
  • The complaint reserves the right to assert other claims Compl. ¶213

U.S. Patent No. 10,713,672

  • Patent Identification: U.S. Patent No. 10,713,672 ("Discovering Neighborhood Clusters and Uses Therefor"), issued July 14, 2020 Compl. ¶43

The Invention Explained

  • Problem Addressed: Similar to the '887 Patent, the '672 Patent addresses the limitations of defining neighborhoods based on static boundaries or subjective perceptions Compl. ¶¶48-49
  • The Patented Solution: The '672 Patent claims a method for determining neighborhood clusters using "statistical inference from a probability distribution based on patterns of check-in time" Compl. ¶50 This allows for the identification of clusters where the mix of venues is "emblematic of one of a predetermined number of temporal check-in pattern types" '672 Patent, abstract The invention can thus group venues not just by who visits them, but by when they are visited, revealing temporal-social patterns in a city '672 Patent, col. 16:20-31
  • Technical Importance: This method introduces a layer of temporal analysis, allowing for the discovery of neighborhood clusters defined by shared daily, weekly, or seasonal rhythms of activity Compl. ¶50

Key Claims at a Glance

  • The complaint asserts independent method claim 21 Compl. ¶302
  • Essential elements of Claim 21 include:
    • Storing, in a computer database system, derived venue check-in data that includes check-in time data.
    • Identifying two or more geographic clusters of venues using "statistical inference from a probability distribution, based on patterns of check-in time in the venue check-in data," such that each cluster is emblematic of a temporal check-in pattern type.
    • Transmitting the identified clusters to an analytics server system.
  • The complaint reserves the right to assert other claims Compl. ¶301

Multi-Patent Capsule: U.S. Patent No. 11,222,349

  • Patent Identification: U.S. Patent No. 11,222,349 ("Discovering Neighborhood Clusters and Uses Therefor"), issued January 11, 2022 Compl. ¶52
  • Technology Synopsis: The '349 Patent describes methods for determining neighborhood clusters by first calculating a "social similarity score between venues" based on "commonalities in the check-in data," and then identifying a cluster based on those scores. It also discloses a method using a "graph representation having nodes representing venues and edges weighted based on the affinity score" Compl. ¶59
  • Asserted Claims: The complaint identifies independent claims 1 and 11 as exemplary Compl. ¶59
  • Accused Features: The complaint alleges that Uber's H3 hexclustering system, which clusters hexagons based on similarities in event data to forecast demand and set prices, infringes the '349 Patent Compl. ¶¶400, 402

Multi-Patent Capsule: U.S. Patent No. 11,935,082

  • Patent Identification: U.S. Patent No. 11,935,082 ("Discovering Neighborhood Clusters and Uses Therefor"), issued March 19, 2024 Compl. ¶61
  • Technology Synopsis: The '082 Patent recites a method that involves calculating both a "spatial proximity" and a "social similarity score" between pairs of venues. These two values are then used to calculate a combined "affinity score," which forms the basis for identifying a neighborhood cluster Compl. ¶68
  • Asserted Claims: The complaint identifies independent claim 1 as exemplary Compl. ¶68
  • Accused Features: The complaint alleges that Uber's hexclustering system infringes by considering both geographic proximity and social similarity (based on user behavior) to calculate an affinity score for clustering hexagons Compl. ¶¶478, 481, 490

III. The Accused Instrumentality

  • Product Identification: The accused instrumentalities are Defendants' "H3 hexagon and hexclustering systems" and the products and services that use them, including the Uber Platform (Mobility, Delivery, and Freight offerings) Compl. ¶207
  • Functionality and Market Context: The complaint alleges that Uber developed its H3 system to overcome problems with earlier, manually-drawn geofences used for surge pricing, which created issues like "surge cliffs" (abrupt price changes over a boundary) and "phantom demand" (mismatch between displayed and actual demand location) Compl. ¶¶153-157 The H3 system divides the world into a grid of nested hexagons, which serve as the fundamental unit for analysis Compl. ¶¶160, 162 A figure in the complaint depicts this system as a hexagonal grid covering the globe Compl. p. 24 Uber allegedly "buckets" event data, such as ride requests, into these hexagonal cells Compl. ¶171 It then forms "hexclusters" by grouping hexagons based on similarities derived from historical and real-time event data Compl. ¶¶186, 196 These hexclusters are used for various applications, most notably to forecast demand and calculate dynamic "surge pricing" by measuring supply and demand within the hexagons Compl. ¶¶189, 192

IV. Analysis of Infringement Allegations

9,846,887 Infringement Allegations

Claim Element (from Independent Claim 23) Alleged Infringing Functionality Complaint Citation Patent Citation
storing, in a computer database system, derived venue check-in data that is based on time-stamped location data captured by a plurality of electronic location sensors ... Uber collects and stores event data, including time-stamped location data from users' mobile devices, in a database. The complaint equates H3 hexagons with "venues." ¶218 col. 4:15-19
generating, by one or more processors ... a check-in intensity vector for each of the multiple venues based on the venue check-in data ... Uber generates metrics for each hexagon based on event data, which corresponds to a "check-in intensity vector," to forecast supply and demand for surge pricing. ¶223 col. 5:49-53
generating, by the one or more processors, elements of a pairwise venue similarity matrix for the multiple venues ... wherein the similarity score for a pair of the venues is determined ... based on at least a measure of the similarity between the check-in intensity vectors ... Uber uses its system to identify similarities between hexagons and generates a similarity score for pairs of hexagons, which are stored in a pairwise similarity matrix. ¶223; ¶225 col. 6:20-23
determining, by the one or more processors, boundaries for two or more geographic clusters of venues in the geographic region, wherein the geographic clusters of venues are determined based on at least the generated elements of the pairwise venue similarity matrix ... Uber clusters H3 hexagons to form "hexclusters" and determines their boundaries based on the calculated similarities in the event data. A visual in the complaint depicts this process of shading hexagons to form clusters Compl. p. 29 ¶229 col. 6:55-61
transmitting, by the host computer system to an analytics server system ... the two or more geographic clusters ... Uber uses an analytics server system to manage the generated hexclusters via an electronic data network. ¶231 col. 5:15-24

10,713,672 Infringement Allegations

Claim Element (from Independent Claim 21) Alleged Infringing Functionality Complaint Citation Patent Citation
storing, in a computer database system, derived venue check-in data based on time-stamped location data ... wherein the check-in data from the venue visitors comprises check-in time data; Uber collects and stores event data from users, which includes time-stamped location data, in a database for multiple hexagons in a geographic region. ¶306 col. 6:66-67
identifying, by a host computer system ... two or more geographic clusters of venues ... using statistical inference from a probability distribution, based on patterns of check-in time in the venue check-in data, such that the mix of venues for each cluster is emblematic of one of a predetermined number of temporal check-in pattern types. Uber uses its computer system to identify "hexclusters" based on similarities in event data. The complaint alleges Uber's forecasting and pricing models use statistical inferences and are designed to account for temporal patterns. The resulting clusters are allegedly emblematic of temporal check-in pattern types. ¶308; ¶316; ¶317 col. 16:20-31
transmitting, by the host computer system to an analytics server system ... the two or more geographic clusters identified by the host computer system ... Uber uses an analytics server system in communication with its host system to manage hexclusters, including updating them with real-time data. A diagram shows the "Cluster Generation" and "Cluster Management" components of the system Compl. p. 57 ¶319; ¶320 col. 5:15-24

Identified Points of Contention

  • Scope Questions: A central dispute may arise over the meaning of "venue." The patents describe venues as discrete, human-centric locations like restaurants or stores '887 Patent, col. 3:30-36 The complaint alleges that Uber's "H3 hexagons"-arbitrary cells in a geographic grid system-are equivalent to "venues" Compl. ¶¶217-218 This raises the question of whether an arbitrary geographic area can be construed as a "venue" under the patent's claim language and specification.
  • Technical Questions: The infringement allegations map complex claim language to high-level descriptions of Uber's system. This suggests potential disputes over functional equivalence. For the '887 Patent, a question is whether Uber's method of measuring supply and demand in a hexagon performs the specific function of generating a "check-in intensity vector" as claimed. For the '672 Patent, a key question will be whether Uber's forecasting algorithms perform "statistical inference from a probability distribution, based on patterns of check-in time" to create clusters "emblematic" of specific "temporal check-in pattern types," as the claim requires, or if they operate on a different technical basis.

V. Key Claim Terms for Construction

  • The Term: "venue"

    • Context and Importance: The viability of the infringement case appears to hinge on construing "venue" to read on Uber's "H3 hexagons." If a hexagon is not a "venue," the entire infringement theory may fail. Practitioners may focus on this term because it represents a potential definitional mismatch between the patent's context and the accused system's architecture.
    • Intrinsic Evidence for Interpretation:
      • Evidence for a Broader Interpretation: The specification provides a non-exhaustive list of examples, concluding with "or any other indoor or outdoor point of interest for which a user might wish to share their presence" '887 Patent, col. 3:33-36 Plaintiff may argue that an H3 hexagon containing user activity is an "outdoor point of interest."
      • Evidence for a Narrower Interpretation: The specification's examples are all discrete, named locations like "restaurants, bars, churches, retail stores, hospitals" '887 Patent, col. 3:30-32 The context is consistently tied to location-based social networking apps like Foursquare, where users check in to specific, human-recognized places, not arbitrary grid cells '887 Patent, col. 3:36-39

  • The Term: "check-in" (and "venue check-in data")

    • Context and Importance: The patents are based on analyzing "check-in" data. The complaint equates this with Uber's "event data," such as ride requests Compl. ¶¶171-172 Whether a ride request from a location constitutes a "check-in" at that location as contemplated by the patent is a critical question for infringement.
    • Intrinsic Evidence for Interpretation:
      • Evidence for a Broader Interpretation: The specification contemplates data sources beyond explicit check-in apps, including "credit card, debit card, gift card, or other purchase data" and "sensor data" '887 Patent, col. 2:49-55 This may support an argument that any data point indicating a user's presence at a location, like a ride request, qualifies as a "check-in."
      • Evidence for a Narrower Interpretation: The primary embodiment described revolves around location-based social networking apps where users actively "check in" to a venue to share their presence '887 Patent, col. 3:5-11 This context of an intentional, user-initiated action to associate with a specific place may support a narrower definition that excludes transient system events like ride requests.

VI. Other Allegations

  • Indirect Infringement: The complaint alleges both induced and contributory infringement. Inducement is based on allegations that Uber instructs and encourages its partners and customers (e.g., through user manuals and technical materials for the Uber Freight app) to use the infringing hexclustering system Compl. ¶¶267-269 Contributory infringement is based on the allegation that Uber's hexclustering system is a material component of the patented method, is especially made for this use, and has no substantial non-infringing uses Compl. ¶¶273-275
  • Willful Infringement: Willfulness is alleged based on knowledge of the patents "at least as of the date of service of this Complaint" Compl. ¶294 Compl. ¶381 Compl. ¶463 Compl. ¶551 The complaint also factually alleges that an Uber engineer contacted the inventors in 2012 to express interest in the underlying "Livehoods" research project, years before the first patent issued Compl. ¶30 This allegation may be used later to support a claim of pre-suit knowledge or willful blindness.

VII. Analyst's Conclusion: Key Questions for the Case

  • A core issue will be one of definitional scope: can the term "venue," rooted in the patent's context of discrete, human-centric locations like restaurants and shops, be construed to cover the arbitrary, algorithmically-defined "H3 hexagons" that form the basis of Uber's accused dynamic pricing system?
  • A key evidentiary question will be one of functional equivalence: does Uber's collection of user "event data" (e.g., ride requests) within a hexagon perform the same function as the "check-in intensity vector" or "patterns of check-in time" required by the patent claims, or is there a fundamental mismatch in the technical operation of the accused system versus the patented methods?
  • A third question will concern knowledge and willfulness: what was the extent of Defendant's knowledge of the patented technology, particularly in light of the complaint's allegation that an Uber engineer contacted the inventors about their "Livehoods" project in 2012, and how might this pre-patent-issuance contact influence the analysis of willful infringement?
Loading Complaint