Autonomous Devices LLC v. Tesla Inc

I. Executive Summary and Procedural Information

• Parties & Counsel:
  • Plaintiff: Autonomous Devices LLC (Florida)
  • Defendant: Tesla, Inc. (Delaware)
  • Plaintiff's Counsel: Shaw Keller LLP
• Case Identification: 1:22-cv-01466, D. Del., 11/07/2022
• Venue Allegations: Venue is asserted in the District of Delaware on the basis that Defendant Tesla, Inc. is a Delaware corporation.
• Core Dispute: Plaintiff alleges that Defendant's autonomous vehicle systems (Autopilot, Full Self-Driving) and AI training supercomputer (Dojo) infringe six patents related to machine learning methods for autonomous device operation based on fleet-learned data and simulation.
• Technical Context: The technology at issue addresses methods for training autonomous systems, particularly vehicles, by learning from a large fleet of devices and using simulation to handle a wide variety of scenarios.
• Key Procedural History: The complaint does not mention prior litigation. However, public records indicate that post-filing, the patentee has filed disclaimers for several asserted claims. For U.S. Patent No. 10,452,974, claims 1, 2, and 18 (among others) were disclaimed. For U.S. Patent No. 11,238,344, claims 1, 2, and 19 (among others) were disclaimed. For U.S. Patent No. 11,055,583, an Inter Partes Review resulted in the cancellation of several claims. These post-grant events will significantly impact the scope and viability of the asserted claims.

Case Timeline

Date	Event
2014-10-01	Tesla begins equipping Model S with Autopilot hardware (HW1)
2016-10-01	Tesla transitions to Autopilot hardware version 2 (HW2)
2016-12-19	Earliest Priority Date for '134 and '585 Patents
2017-08-01	Tesla announces hardware version 2.5 (HW2.5)
2017-11-21	Earliest Priority Date for '449 and '583 Patents
2018-10-05	Tesla releases Software Version 9.0; alleged infringement begins
2018-10-16	U.S. Patent No. 10,102,449 Issues
2018-11-02	Earliest Priority Date for '974 and '344 Patents
2019-03-01	Tesla releases hardware version 3 (HW3)
2019-10-22	U.S. Patent No. 10,452,974 Issues
2020-03-31	U.S. Patent No. 10,607,134 Issues
2020-08-01	Tesla announces training its car computers in the "Dojo" simulation
2021-07-06	U.S. Patent No. 11,055,583 Issues
2021-09-07	U.S. Patent No. 11,113,585 Issues
2022-02-01	U.S. Patent No. 11,238,344 Issues
2022-11-07	Complaint Filed

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

U.S. Patent No. 10,452,974 - "Artificially intelligent systems, devices, and methods for learning and/or using a device's circumstances for autonomous device operation"

The Invention Explained

• Problem Addressed: The patent asserts that prior art autonomous systems were limited because they relied on pre-coded behaviors and could not learn from the vast array of real-world scenarios or share knowledge gained by one device with others (Compl. ¶42; Compl. ¶43, Compl. ¶¶col. 1:26-35).
• The Patented Solution: The invention describes a method where a "first device," often operated by a human, correlates its environmental circumstances (sensed data) with the operations performed (user actions) to create a knowledgebase of circumstance-instruction pairs Compl. ¶44 A "second device" can then sense its own, new circumstances, find a partial match in the knowledgebase, and autonomously execute the learned instructions, enabling a "fleet learning" capability '974 Patent, abstract '974 Patent, col. 2:38-61 Figure 38, reproduced in the complaint, illustrates a vehicle detecting objects like people and other vehicles in its surroundings Compl. p. 18
• Technical Importance: This approach allows for the creation of more robust autonomous systems that can improve over time by collectively learning from the experiences of an entire fleet, rather than being limited to pre-programmed logic Compl. ¶¶48-49

Key Claims at a Glance

• The complaint asserts independent claims 1 and 18, and dependent claim 14 Compl. ¶95 Note: A subsequent disclaimer has canceled claims 1 and 18.
• For illustrative purposes, the elements of the originally asserted independent claim 1 include:
  • Accessing a memory with a knowledgebase of learned correlations between a "first circumstance representation" and "instruction sets for operating a first device," where the learning involves a "user."
  • Generating or receiving a "third circumstance representation" from a "second device."
  • "Anticipating" the instruction sets from the knowledgebase based on a partial match between the first and third circumstance representations.
  • "Executing" the anticipated instruction sets to cause the first or second device to "autonomously" perform an operation.

U.S. Patent No. 11,238,344 - "Artificially intelligent systems, devices, and methods for learning and/or using a device's circumstances for autonomous device operation"

The Invention Explained

• Problem Addressed: As with its parent '974 patent, the '344 patent addresses the limitations of pre-programmed autonomous systems that cannot learn from or share real-world experiences Compl. ¶42 '344 Patent, col. 1:26-35
• The Patented Solution: The '344 patent claims a system embodying the same fleet-learning concept. It describes one or more processors configured to access a knowledgebase of learned circumstance-instruction correlations from a "first device" and apply them to enable autonomous operation in a "second device" upon detecting a matching circumstance '344 Patent, abstract '344 Patent, col. 2:38-61
• Technical Importance: The technology provides a system-level architecture for implementing a scalable, learning-based approach to autonomous control, moving beyond the limitations of isolated, rule-based devices Compl. ¶¶51-52

Key Claims at a Glance

• The complaint asserts independent claim 1 and dependent claim 3 Compl. ¶102 Note: A subsequent disclaimer has canceled claim 1.
• For illustrative purposes, the elements of the originally asserted independent claim 1 include:
  • One or more processors configured to access a memory with a knowledgebase of "circumstance representation" and "instruction set" correlations learned from a "first device" operated by a "user."
  • The processors are further configured to receive a "second circumstance representation" from the first or a "second device."
  • The processors anticipate the instruction sets based on a partial match between the circumstance representations.
  • The processors cause the first or second device to perform an operation defined by the anticipated instruction sets.

Multi-Patent Capsule: U.S. Patent No. 10,102,449

• Patent Identification: U.S. Patent No. 10,102,449 ("Devices, systems, and methods for use in automation"), issued October 16, 2018 Compl. ¶54
• Technology Synopsis: This patent is directed to an autonomous system that learns by correlating "digital pictures" with instruction sets. When the system encounters a new digital picture, it matches it against a knowledgebase of previously learned picture-instruction pairs to find a corresponding operation to execute autonomously Compl. ¶¶58-60
• Asserted Claims: Independent claims 1 and 17 Compl. ¶109
• Accused Features: Tesla's Autopilot, Enhanced Autopilot, and FSD systems, which allegedly use camera-captured digital pictures and fleet-learned knowledge to make driving decisions (Compl. ¶¶83; Compl. ¶110).

Multi-Patent Capsule: U.S. Patent No. 11,055,583

• Patent Identification: U.S. Patent No. 11,055,583 ("Machine learning for computing enabled systems and/or devices"), issued July 6, 2021 Compl. ¶55
• Technology Synopsis: This patent continues the "Digital Picture" theme, focusing on a system that receives new digital pictures, determines a match with previously stored pictures, and in response, causes a device to perform an operation defined by an instruction set learned from a first device Compl. ¶58
• Asserted Claims: Independent claim 4 Compl. ¶116
• Accused Features: Tesla's vision-based autonomous driving systems that allegedly match current camera views to a learned knowledgebase of driving scenarios and actions (Compl. ¶¶83; Compl. ¶116).

Multi-Patent Capsule: U.S. Patent No. 10,607,134

• Patent Identification: U.S. Patent No. 10,607,134 ("Artificially intelligent systems, devices, and methods for learning and/or using an avatar's circumstances for autonomous avatar operation"), issued March 31, 2020 Compl. ¶69
• Technology Synopsis: This patent adapts the core learning technology to a simulated environment. It describes a system where an "avatar" (e.g., a simulated vehicle) learns to perform operations by matching its current simulated circumstances (object representations) against a knowledgebase of learned circumstance-instruction correlations Compl. ¶¶72-73
• Asserted Claims: Independent claim 1 Compl. ¶122
• Accused Features: Tesla's "Dojo" supercomputer and associated simulation software, which allegedly train and validate autonomous driving behavior using simulated avatars and environments (Compl. ¶¶84; Compl. ¶123).

Multi-Patent Capsule: U.S. Patent No. 11,113,585

• Patent Identification: U.S. Patent No. 11,113,585 ("Artificially intelligent systems, devices, and methods for learning and/or using visual surrounding for autonomous object operation"), issued September 7, 2021 Compl. ¶70
• Technology Synopsis: This patent also focuses on simulation, describing a system that correlates digital pictures from a simulated environment with instruction sets. When a new digital picture is generated in the simulation, the system finds a match in a learned knowledgebase and causes a simulated object to autonomously perform the corresponding operation Compl. ¶73
• Asserted Claims: Independent claim 1 Compl. ¶129
• Accused Features: The Dojo supercomputer and Tesla's simulation software, which are alleged to use real-world and synthetic digital pictures to train the driving AI Compl. ¶84 Compl. ¶85

III. The Accused Instrumentality

Product Identification

• The complaint identifies two main categories of accused instrumentalities:
  1. "Tesla Fleet Vehicles": This includes at least Tesla Models S, 3, X, and Y equipped with "Software Version 9.0 and beyond," which feature Autopilot, Enhanced Autopilot, and/or Full Self-Driving (FSD) capabilities Compl. ¶11 fn.1 Compl. ¶83
  2. "Dojo Supercomputer" and "Infringing Simulation Software": This refers to Tesla's AI training hardware and the software used to simulate driving scenarios for training and testing its autonomous systems Compl. ¶84

Functionality and Market Context

• The complaint alleges that Tesla's vehicles use an array of sensors, primarily cameras, to perceive their surroundings Compl. ¶32 This data is used to enable autonomous driving features. Central to the plaintiff's theory is the concept of "fleet learning," where data from the entire fleet of customer vehicles is collected and used to train the system's neural networks, with improvements then distributed back to the vehicles via over-the-air software updates Compl. ¶17 Compl. Ex. I, p. 2 The Dojo supercomputer is alleged to be the platform where this large-scale training occurs, using both real-world video and simulated scenarios to refine the AI Compl. ¶85 The complaint provides a screenshot from a Tesla presentation titled "5. Scenario Reconstruction" to support its allegations regarding the use of simulation Compl. ¶84 This visual depicts a pipeline from "Real World City" to "Auto-Labeled Reconstruction" to "Recreated Synthetic World," which plaintiff alleges demonstrates the accused simulation functionality Compl. ¶84

IV. Analysis of Infringement Allegations

Given that the lead independent claims of both the '974 and '344 patents have been disclaimed post-filing, the following chart summaries are presented for informational purposes based on the original complaint's allegations. The viability of these infringement theories is now in question.

'974 Patent Infringement Allegations

Claim Element (from Independent Claim 1)	Alleged Infringing Functionality	Complaint Citation	Patent Citation
[1a] accessing a memory that stores at least a knowledgebase that includes... a first correlation... wherein at least a portion of the first correlation... are learned in a learning process that includes operating the first device at least partially by a user;	A first Tesla vehicle, operated by a human driver, learns driving instructions. This creates a correlation between a circumstance (e.g., pedestrian in road) and an instruction (e.g., brake) which is stored in a knowledgebase on the Dojo supercomputer or within the fleet.	Ex. I, p. 7	col. 2:38-48
[1b] generating or receiving a third circumstance representation, wherein the third circumstance representation represents a third circumstance detected at least in part by one or more sensors of a second device;	A second Tesla vehicle's cameras and sensors detect a new circumstance, such as a pedestrian in front of the vehicle, generating a new circumstance representation.	Ex. I, p. 13	col. 16:66-17:2
[1c] anticipating the first one or more instruction sets... based on at least a partial match between the third circumstance representation and the first circumstance representation;	The processor of the second vehicle anticipates a set of driving instructions by matching the newly detected circumstance against the previously learned correlations stored in the knowledgebase.	Ex. I, p. 15	col. 17:15-28
[1d] ...executing the first one or more instruction sets... wherein the first device or a second device autonomously performs one or more operations...	The second vehicle autonomously executes the anticipated instruction (e.g., applying the brakes), thereby performing an operation learned from the first vehicle.	Ex. I, p. 18	col. 17:29-41

'344 Patent Infringement Allegations

Claim Element (from Independent Claim 1)	Alleged Infringing Functionality	Complaint Citation	Patent Citation
[1a] One or more processors configured to perform at least: accessing a memory that stores at least a knowledgebase that includes... a first correlation... wherein at least a portion of the first one or more instruction sets... is learned in a learning process that includes operating the first device at least partially by a user;	The processors in Tesla vehicles and/or the Dojo supercomputer are configured to access a memory storing a knowledgebase of driving instructions learned from human drivers across the fleet.	Ex. J, pp. 7-9	col. 2:38-51
[1b] generating or receiving a second circumstance representation, wherein the second circumstance representation represents a second circumstance detected at least in part by... one or more sensors of a second device;	The processor in a second Tesla vehicle is configured to receive a representation of the vehicle's current surroundings from its cameras and other sensors.	Ex. J, p. 13	col. 17:2-8
[1c] anticipating the first one or more instruction sets for operating the first device based on at least a partial match between the second circumstance representation and the first circumstance representation;	The processor is configured to anticipate learned driving instructions by comparing the current circumstance representation to the stored knowledgebase to find a partial match.	Ex. J, p. 15	col. 17:20-33
[1d] at least in response to the anticipating, causing the first device or the second device to perform one or more operations defined by the first one or more instruction sets...	The processor is configured to cause the vehicle to autonomously execute the matched, learned instruction sets (e.g., steering, braking).	Ex. J, p. 18	col. 17:34-42

• Identified Points of Contention:
  • Scope Questions: A primary question may concern the "first device" / "second device" architecture. The patents describe learning from a first device and applying it to a second. The complaint alleges the "first device" is the entire fleet of human-driven Teslas and the "second device" is any single Tesla operating autonomously Compl. Ex. I, p. 2 A court may need to determine if this distributed, aggregated learning model falls within the scope of the claims, which could be read to imply a more direct device-to-device knowledge transfer.
  • Technical Questions: The meaning of the "learning process" will be critical. The complaint relies on Tesla's public statements that human drivers are "training the neural net" (Compl. Ex. I, p. 7). A potential dispute is whether this passive data collection from everyday driving constitutes the specific "learning" of a "correlation" between a "circumstance representation" and an "instruction set" as required by the claims, or if the patent requires a more structured, supervised training process.

V. Key Claim Terms for Construction

• The Term: "circumstance representation"

• Context and Importance: This term defines the input data for the patented system. Its scope is central to whether Tesla's sensor data infringes. The complaint equates this term with "object representation" Compl. ¶41 fn. 26 and alleges Tesla's system uses it Compl. ¶46 Practitioners may focus on whether this term requires a specific data structure or broadly covers any sensor-derived model of the environment.

• Intrinsic Evidence for Interpretation:

  • Evidence for a Broader Interpretation: The specification states that the learning can be based on "any information on a device's circumstances" '974 Patent, col. 58:1-2, and that a circumstance representation may include "one or more objects detected by the sensor" '974 Patent, col. 2:50-51, which could support a broad interpretation covering raw or processed sensor data.
  • Evidence for a Narrower Interpretation: The patent frequently illustrates the concept with a specific, structured "Collection of Object Representations" containing discrete fields like "Type," "Distance," and "Bearing" '974 Patent, FIG. 4B '974 Patent, col. 80:5-24 A defendant may argue that the term should be limited to such structured data, as opposed to the vector-space representations common in modern neural networks.

• The Term: "learning process that includes operating the first device at least partially by a user"

• Context and Importance: This term is the foundation of the plaintiff's "fleet learning" theory of infringement. The definition will determine whether the routine operation of Tesla vehicles by their owners qualifies as the claimed "learning process."

• Intrinsic Evidence for Interpretation:

  • Evidence for a Broader Interpretation: The plain language "at least partially by a user" appears broad. The specification contemplates learning from a user's "knowledge, style, or methodology" '974 Patent, col. 8:1-2, which could encompass general driving behavior. The complaint points to Tesla's own description of its "data engine" and "shadow mode" as evidence of this process (Compl. Ex. I, p. 7).
  • Evidence for a Narrower Interpretation: The term "learning process" could be construed to require a more active and intentional training phase, rather than passive data aggregation from millions of unrelated drivers. The specification's focus on a "first device" and a "second device" could suggest a more contained learning cycle, which a defendant may argue is distinct from Tesla's large-scale, continuous model refinement.

VI. Other Allegations

• Indirect Infringement: The complaint alleges induced infringement under 35 U.S.C. § 271(b). It asserts that Tesla encourages its customers to infringe by providing instructions, marketing, and user interfaces (such as the in-car display and mobile app) that facilitate and promote the use of the accused Autopilot and FSD features (Compl. ¶12; Compl. ¶13; Compl. ¶14; Compl. ¶15; Compl. ¶16; Compl. ¶17; Compl. ¶18; Compl. ¶19; Compl. ¶20; Compl. ¶21; Compl. ¶22; Compl. ¶23; Compl. ¶24; Compl. ¶25; Compl. ¶26; Compl. ¶27; Compl. ¶28; Compl. ¶29; Compl. ¶30; Compl. ¶31; Compl. ¶32; Compl. ¶33; Compl. ¶34; Compl. ¶35; Compl. ¶36; Compl. ¶37; Compl. ¶38; Compl. ¶39; Compl. ¶40; Compl. ¶41; Compl. ¶42; Compl. ¶43; Compl. ¶44; Compl. ¶45; Compl. ¶46; Compl. ¶47; Compl. ¶48; Compl. ¶49; Compl. ¶50; Compl. ¶51; Compl. ¶52; Compl. ¶53; Compl. ¶54; Compl. ¶55; Compl. ¶56; Compl. ¶57; Compl. ¶58; Compl. ¶59; Compl. ¶60; Compl. ¶61; Compl. ¶62; Compl. ¶63; Compl. ¶64; Compl. ¶65; Compl. ¶66; Compl. ¶67; Compl. ¶68; Compl. ¶69; Compl. ¶70; Compl. ¶71; Compl. ¶72; Compl. ¶73; Compl. ¶74; Compl. ¶75; Compl. ¶76; Compl. ¶77; Compl. ¶78; Compl. ¶79; Compl. ¶80; Compl. ¶81; Compl. ¶82; Compl. ¶83; Compl. ¶84; Compl. ¶85; Compl. ¶86; Compl. ¶87; Compl. ¶88; Compl. ¶89; Compl. ¶90). The complaint includes screenshots of the Tesla app showing options to purchase "Enhanced Autopilot" and "Full Self-Driving Capability" as evidence of this encouragement Compl. p. 40
• Willful Infringement: The complaint alleges that Tesla has had "actual and/or constructive knowledge" of the asserted patents since at least the filing of the complaint and that its continued infringement is willful (Compl. ¶¶92; 97). The allegations primarily point to post-suit knowledge as the basis for willfulness.

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

1. A central issue will be one of architectural scope: Can the patents' "first device/second device" learning architecture be construed to cover Tesla's distributed fleet-learning model, where knowledge is aggregated from millions of vehicles into a central AI model and then redeployed, or is there a fundamental mismatch between the claimed architecture and the accused system?

2. A key question of claim construction will be the definition of "learning": Does the claimed "learning process" involving a "user" simply require passive data collection from drivers, as the plaintiff alleges, or does it require a more specific, structured training activity where discrete environmental circumstances are explicitly correlated with user-provided instructions, as the patent's embodiments might suggest?

3. A threshold legal question will be the effect of post-grant proceedings: Given that the plaintiff has disclaimed the lead independent claims of the '974 and '344 patents after filing the complaint, a primary focus will be on whether the infringement allegations can be sustained under any remaining asserted dependent claims and how the prosecution history of these disclaimers might limit the scope of those remaining claims.