DCT

2:25-cv-00742

Perceptive Automata LLC v. Tesla Inc

Log In
Key Events
Amended Complaint

I. Executive Summary and Procedural Information

  • Parties & Counsel:
  • Case Identification: 2:25-cv-00742, E.D. Tex., 11/14/2025
  • Venue Allegations: Plaintiff alleges venue is proper in the Eastern District of Texas because Tesla has regular and established places of business within the district and has committed acts of infringement there, including the sale and use of vehicles with the accused Full Self-Driving (FSD) technology.
  • Core Dispute: Plaintiff alleges that Defendant’s FSD technology, and the datacenter systems used to train it, infringe five patents related to using machine learning models to predict the behavior and intent of humans for autonomous vehicle navigation.
  • Technical Context: The patents address a core challenge in autonomous driving: moving beyond simple motion-vector prediction to understand and anticipate the nuanced "state of mind" of pedestrians, cyclists, and other drivers.
  • Key Procedural History: This First Amended Complaint follows an original complaint and, notably, references a prior motion to dismiss filed by Tesla. Plaintiff alleges that in its motion, Tesla did not dispute the claimed priority dates for several of the asserted patents. The complaint also alleges that Tesla was aware of Plaintiff's patent portfolio prior to the lawsuit due to citations made during Tesla's own patent prosecution activities.

Case Timeline

Date Event
2017-07-05 Priority Date for U.S. Patent Nos. 10,614,344; 11,126,889; and 11,753,046
2019-01-30 Priority Date for U.S. Patent No. 11,520,346
2019-02-06 Priority Date for U.S. Patent No. 11,467,579
2020-04-07 U.S. Patent No. 10,614,344 Issues
2020-10-XX Tesla FSD Beta Software Released (as alleged)
2021-09-21 U.S. Patent No. 11,126,889 Issues
2022-10-11 U.S. Patent No. 11,467,579 Issues
2022-12-06 U.S. Patent No. 11,520,346 Issues
2023-09-12 U.S. Patent No. 11,753,046 Issues
2024-XX-XX Tesla re-incorporates from Delaware to Texas (as alleged)
2025-11-14 First Amended Complaint Filed

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

U.S. Patent No. 10,614,344 - "System and Method of Predicting Human Interaction with Vehicles"

  • Patent Identification: U.S. Patent No. 10,614,344, "System and Method of Predicting Human Interaction with Vehicles," issued April 7, 2020.

The Invention Explained

  • Problem Addressed: The patent asserts that conventional autonomous driving systems fail to adequately predict human behavior because they rely on simple "motion vectors" (i.e., past and current movement) rather than understanding intent, a deficiency particularly acute in complex urban environments (’344 Patent, col. 1:32-49).
  • The Patented Solution: The invention proposes a computer-implemented method where a supervised learning model is trained on data derived from human observers. These observers are shown road scenarios and provide responses about predicted behavior. The system then applies this trained model to live sensor data from a vehicle to generate a "statistical summary" of expected human responses, which is in turn used to control the vehicle's operation (’344 Patent, Abstract; col. 2:4-16). Figure 2A illustrates this process, from receiving video data to applying a trained model to predict user behavior (’344 Patent, FIG. 2A).
  • Technical Importance: This approach seeks to imbue an autonomous system with a predictive capability analogous to human intuition by training it on aggregated human judgments, moving beyond purely kinematic analysis to infer the "state of mind" of road users (’344 Patent, col. 1:21-31; col. 7:35-39).

Key Claims at a Glance

  • The complaint asserts independent claims 1 (a method) and 19 (a computing system) (Compl. ¶22).
  • The core elements of independent claim 1 include:
    • receiving sensor data displaying an object on the road from an autonomous vehicle;
    • providing the sensor data to a supervised learning model configured to predict an output statistical summary characterizing a distribution of user responses;
    • executing the model to generate said statistical summary data; and
    • controlling the vehicle's operation based on the generated statistical summary data.
  • Claim 19 recites a computing system with a processor and memory configured to perform substantially the same steps as method claim 1.
  • The complaint expressly reserves the right to assert additional claims (Compl. ¶22, n.1).

U.S. Patent No. 11,126,889 - "Machine Learning Based Prediction of Human Interactions with Autonomous Vehicles"

  • Patent Identification: U.S. Patent No. 11,126,889, "Machine Learning Based Prediction of Human Interactions with Autonomous Vehicles," issued September 21, 2021.

The Invention Explained

  • Problem Addressed: Like its parent, the ’344 patent, this patent addresses the shortcomings of autonomous systems that cannot "adequately predict the behavior of people, especially in urban environments" by relying on simple motion prediction (’889 Patent, col. 1:39-53).
  • The Patented Solution: The invention outlines a system and method that separates the training and deployment phases. A training dataset is first created by collecting images, gathering "user responses describing a state of mind" from human observers viewing those images, and generating "summary statistics" from these responses. A supervised learning model is then trained on this dataset. Subsequently, an autonomous vehicle uses this trained model to receive a new, live image and predict the state of mind of a road user within it to inform control decisions (’889 Patent, Abstract; col. 2:5-24).
  • Technical Importance: This patent details a systematic approach for creating and deploying a model based on human psychological assessments ("state of mind"), establishing a framework for translating subjective human intuition into actionable data for an autonomous vehicle.

Key Claims at a Glance

  • The complaint asserts independent claims 1 (a system) and 2 (a method) (Compl. ¶84).
  • The core elements of independent claim 2 include:
    • receiving a plurality of images of road scenes;
    • receiving a plurality of user responses describing a "state of mind" of a road user in the images;
    • generating a training dataset with summary statistics of the user responses;
    • training a supervised learning model on the dataset to predict such summary statistics;
    • in an autonomous vehicle, receiving a new image and using the model to predict the road user's state of mind; and
    • controlling the vehicle based on the prediction.
  • Claim 1 recites a computer system configured to perform substantially the same steps.
  • The complaint expressly reserves the right to assert additional claims (Compl. ¶84, n.1).

U.S. Patent No. 11,467,579 - "Probabilistic Neural Network for Predicting Hidden Context of Traffic Entities for Autonomous Vehicles"

  • Patent Identification: U.S. Patent No. 11,467,579, "Probabilistic Neural Network for Predicting Hidden Context of Traffic Entities for Autonomous Vehicles," issued October 11, 2022.
  • Technology Synopsis: This patent discloses using a "probabilistic neural network" that takes a traffic image as input and generates an output representing the "hidden context" (e.g., intent) of a traffic entity. Critically, the system also determines a "measure of uncertainty" for this prediction, and the vehicle uses both the predicted context and its associated uncertainty to navigate, for example, by adjusting its safe following distance (’579 Patent, Abstract).
  • Asserted Claims: Claims 1 and 20 (Compl. ¶148).
  • Accused Features: The complaint alleges Tesla’s FSD software functions as the claimed probabilistic neural network, generating outputs representing hidden context for traffic entities and determining a measure of uncertainty to navigate the vehicle, for example, by deciding to pull over for an oncoming car (Compl. ¶¶157-161).

U.S. Patent No. 11,520,346 - "Navigating Autonomous Vehicles Based on Modulation of a World Model Representing Traffic Entities"

  • Patent Identification: U.S. Patent No. 11,520,346, "Navigating Autonomous Vehicles Based on Modulation of a World Model Representing Traffic Entities," issued December 6, 2022.
  • Technology Synopsis: This invention describes a method of "modulating" an autonomous vehicle's "world model" (e.g., a point cloud map of its surroundings) using predicted hidden context. After identifying a traffic entity, the system predicts its hidden context and modifies the region of the world model where the entity is expected to travel. The vehicle then navigates based on this dynamically modified, context-aware map (’346 Patent, Abstract).
  • Asserted Claims: Claims 1 and 19 (Compl. ¶209).
  • Accused Features: Tesla's FSD system is accused of generating a point cloud representation of its surroundings, using a model to determine the "hidden context" of traffic entities, modifying the expected path of those entities based on that context, and then navigating the vehicle in response to the modified path (Compl. ¶¶216-223).

U.S. Patent No. 11,753,046 - "System and Method of Predicting Human Interaction with Vehicles"

  • Patent Identification: U.S. Patent No. 11,753,046, "System and Method of Predicting Human Interaction with Vehicles," issued September 12, 2023.
  • Technology Synopsis: This patent, from the same family as the '344 and '889 patents, focuses on the method of creating the training data and model. It claims a method of storing images, sending them to human observers with a question about a user's "state of mind," receiving their responses, generating and storing summary statistics, and using this data to train a predictive model. The final step is executing the trained model to predict a user's state of mind in a new image (’046 Patent, Claim 1).
  • Asserted Claims: Claims 1 and 15 (Compl. ¶271).
  • Accused Features: The complaint targets Tesla's "Data Engine" and associated datacenters (e.g., Dojo, Cortex) as the infringing systems. It alleges these systems store video from the Tesla fleet and use a large team of human labelers to provide judgments about road user "state of mind," thereby generating the training data used to create and update the FSD model (Compl. ¶¶276-282).

III. The Accused Instrumentality

Product Identification

The accused instrumentalities are Tesla vehicles equipped with Full Self-Driving (FSD) hardware and software, including the Model Y and Cybertruck, as well as the supporting backend infrastructure, such as the "Cortex" and "Dojo" datacenters used for training the FSD neural networks (Compl. ¶¶15, 16, 22, 90).

Functionality and Market Context

The complaint alleges that Tesla's FSD system utilizes an onboard computer and an array of cameras to perceive its environment (Compl. ¶¶28, 159). This system is powered by neural networks trained on video data collected from Tesla's customer fleet and processed by over 1,000 in-house data labelers (Compl. ¶¶29, 91). The complaint alleges the system's "Data Engine" uses this labeled data to train models that predict the "state of mind" and "hidden context" of other road users, allowing the vehicle to make control decisions such as yielding to a pedestrian or navigating around other cars (Compl. ¶¶29, 31, 92). A visual from a Tesla AI Day presentation shows the "Data Engine" architecture, which takes data from the "Tesla Fleet" and uses "Simulation + Auto Label + Human Label" to create a training set (Compl. ¶276, p. 191). This technology is fundamental to Tesla’s marketed Autopilot and FSD features and its stated goal of creating an autonomous "robotaxi" service (Compl. ¶17).

IV. Analysis of Infringement Allegations

U.S. Patent No. 10,614,344 Infringement Allegations

Claim Element (from Independent Claim 1) Alleged Infringing Functionality Complaint Citation Patent Citation
receiving, by a computing device associated with an autonomous vehicle operating on a road, sensor data captured by a sensor installed on the autonomous vehicle, the sensor data displaying an object on the road Tesla's onboard computer receives data from its camera sensors, which capture images of objects on the road, such as other vehicles and pedestrians. A visual from Tesla's marketing materials shows the array of cameras on a Model Y (Compl. ¶28, p. 18). ¶28 col. 5:4-7
providing the sensor data as input to a supervised learning based model, the supervised learning based model configured to ... predict an output statistical summary characterizing a distribution of user responses expected to be received responsive to presenting the sensor data to a plurality of users... The FSD software, alleged to be a supervised learning model, receives sensor data and creates "goal candidates" that allegedly "correspond to a probability mask derived from human demonstration," which the complaint equates to a statistical summary of user responses. ¶29 col. 6:1-11
executing the trained supervised learning based model to generate a statistical summary data characterizing a distribution of user responses... The FSD software allegedly executes the model to generate outputs like an "Intervention Likelihood" and inputs for a "Human-like Discriminator," which are alleged to characterize a distribution of expected human driver responses. A diagram from a Tesla presentation illustrates these components (Compl. ¶30, p. 23). ¶30 col. 7:5-8
controlling the operation of the autonomous vehicle on the road based on the generated statistical summary data The FSD software uses its predictive outputs to make control decisions, such as determining whether to "Assert or Yield to the pedestrian" based on an "Interaction Search" analysis. A visual from Tesla's AI Day depicts this decision-making process (Compl. ¶31, p. 25). ¶31 col. 7:35-39
  • Identified Points of Contention:
    • Scope Questions: A central question for the court will be whether Tesla's model, trained on "human demonstrations," predicts a "statistical summary characterizing a distribution of user responses" as required by the claim. The dispute may focus on whether the model's output (e.g., a "probability mask") is equivalent to the claimed "statistical summary," or if the claim requires a more formal statistical process based on explicit user survey-style responses.
    • Technical Questions: The infringement analysis will likely require evidence of how Tesla's "Human-like Discriminator" technically operates and whether it generates a "statistical summary" as its direct output, or if it produces a different type of output (e.g., a simple trajectory score) that the plaintiff argues is equivalent.

U.S. Patent No. 11,126,889 Infringement Allegations

Claim Element (from Independent Claim 2) Alleged Infringing Functionality Complaint Citation Patent Citation
receiving a plurality of images displaying road scenes captured by one or more vehicles Tesla's datacenters allegedly receive millions of video clips captured by vehicles in its customer fleet, as depicted in its "Data Engine" diagram (Compl. ¶90, p. 58). ¶90 col. 5:12-15
receiving a plurality of user responses, each user response describing a state of mind of a road user displayed in one or more images Tesla is alleged to employ over 1,000 human data labelers who provide labels describing the "state of mind" of road users, such as selecting "stopped_traffic" from a list of options. A screenshot shows the labeling interface presented to these users (Compl. ¶91, p. 59). ¶91 col. 6:4-7
generating a training dataset comprising summary statistics of user responses describing the state of minds of road users... Tesla's "Data Engine" is alleged to process these labels to generate a training dataset that incorporates summary statistics, which is used to refine the model's accuracy over time. A chart shows the growth in the "Training set" size (Compl. ¶92, p. 60). ¶92 col. 6:8-18
training, using the training dataset, a supervised learning based model configured to predict summary statistics describing a state of mind of a road user... Tesla is alleged to use this training data to train the FSD software, which is a supervised learning model that predicts actions based on the "possible state of mind of road users." ¶93 col. 6:44-51
receiving, by an autonomous vehicle, a new image captured by a camera of the autonomous vehicle, the new image of a scene including a road user An operating Tesla vehicle receives new, live images from its cameras showing road users like pedestrians. ¶94 col. 7:12-15
predicting, by the autonomous vehicle, using the supervised learning based model, summary statistics describing a state of mind of the road user in the new image The vehicle's FSD software allegedly uses the trained model to make predictions about the road user's state of mind, such as whether a pedestrian is likely to cross the road rather than stand still. ¶95 col. 7:26-31
controlling the autonomous vehicle based on the prediction of the supervised learning based model. Based on its prediction, the FSD software controls the vehicle, for example, by stopping to allow a pedestrian to cross the road without human intervention. ¶96 col. 7:35-44
  • Identified Points of Contention:
    • Scope Questions: The definition of "state of mind" will be a critical issue. The court may need to decide whether objective labels like "stopped_traffic" or "parked" (Compl. ¶91) constitute a description of a "state of mind," or if the term requires a more subjective or psychological assessment (e.g., confused, aware, aggressive).
    • Technical Questions: A key evidentiary question may be how Tesla's backend "Data Engine" processes the inputs from its human labelers. Specifically, does it "generat[e] a training dataset comprising summary statistics of user responses," or does it use the labels in a more direct, non-statistical manner for training?

V. Key Claim Terms for Construction

For U.S. Patent No. 10,614,344

  • The Term: "statistical summary characterizing a distribution of user responses"
  • Context and Importance: This term defines the specific nature of the data used to control the vehicle and is the inventive core of the claim. The infringement analysis depends on whether Tesla's predictive outputs, which the complaint alleges are "probability mask[s] derived from human demonstration" (Compl. ¶29), meet this definition. Practitioners may focus on this term because it distinguishes the invention from models trained on simple, non-probabilistic outcomes.
  • Intrinsic Evidence for Interpretation:
    • Evidence for a Broader Interpretation: The specification describes statistics that may "characterize the aggregate responses of multiple human observers," giving the example of categorizing "how many human observers believe that the pedestrian will stop" versus cross the street ('344 Patent, col. 6:2-11). This language may support an interpretation that covers any data reflecting a probabilistic aggregation of human judgments.
    • Evidence for a Narrower Interpretation: The claims and specification also list specific statistical metrics, including "a central tendency, a variance, a skew, a kurtosis, a scale, and a histogram" ('344 Patent, Claim 6; col. 2:44-46). This could support a narrower construction requiring the generation of one of these formal statistical measures, not just a general likelihood score.

For U.S. Patent No. 11,126,889

  • The Term: "state of mind"
  • Context and Importance: This term is what the claimed invention trains its model to predict. The viability of the infringement allegation hinges on whether the labels provided by Tesla's employees (e.g., selecting "stopped_traffic") can be construed as describing a "state of mind" (Compl. ¶91).
  • Intrinsic Evidence for Interpretation:
    • Evidence for a Broader Interpretation: The specification states that predictions of the "state of mind of road users... can then be used to improve the performance of autonomous vehicles by allowing them to make judgments about the future behavior" ('889 Patent, col. 8:36-44). This suggests a broad, functional definition tied to any internal state that informs future behavior.
    • Evidence for a Narrower Interpretation: The specification provides specific examples of what constitutes a state of mind, including "intention, awareness, personality, state of consciousness, level of tiredness, aggressiveness, enthusiasm, thoughtfulness or another characteristic of the internal mental state" ('889 Patent, col. 9:39-43). A party could argue the term is limited to these or similar subjective, cognitive states, rather than objective, observable situations like a stopped vehicle.

VI. Other Allegations

  • Indirect Infringement: Plaintiff alleges that Tesla induces its customers to infringe by providing FSD software along with instructions, user manuals, and marketing that direct and encourage its use (Compl. ¶¶76-77). The complaint cites an alleged email from Elon Musk making it "mandatory" for employees to give FSD demonstrations to prospective buyers as an affirmative act of inducement (Compl. ¶77).
  • Willful Infringement: The complaint alleges willful infringement based on both pre-suit and post-suit knowledge. Pre-suit knowledge is alleged based on Tesla's alleged citations to Perceptive's patents during its own patent prosecution activities (Compl. ¶75). The complaint asserts that despite this knowledge and notice from the original complaint, Tesla "has nevertheless continued its infringing conduct and disregarded an objectively high likelihood of infringement" (Compl. ¶80).

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

  • A core issue will be one of definitional scope: can terms rooted in human psychology, such as "state of mind" and "user responses," be construed to cover the arguably more objective data-labeling process that Tesla allegedly employs (e.g., categorizing a vehicle as "stopped_traffic")? The case may turn on whether the patents claim a system trained on subjective human judgments versus a system trained on objective event labels provided by humans.
  • A key evidentiary question will be one of technical implementation: what evidence can be presented from within Tesla's proprietary "Data Engine" and FSD software to demonstrate that they technically perform the specific steps of the asserted claims? The dispute will likely involve a deep dive into whether Tesla's system generates and uses "summary statistics" in the manner claimed, or whether it utilizes a fundamentally different machine learning architecture that achieves a similar result through a non-infringing method.