Perceptive Automata LLC v. Tesla Inc

I. Executive Summary and Procedural Information

• Parties & Counsel:
  • Plaintiff: Perceptive Automata LLC (Texas)
  • Defendant: Tesla, Inc. (Texas)
  • Plaintiff’s Counsel: Nelson Bumgardner Conroy PC
• Case Identification: 7:25-cv-00594, W.D. Tex., 03/05/2026
• Venue Allegations: Plaintiff alleges venue is proper because Defendant Tesla has regular and established places of business within the district, including its "Gigafactory" and "Global Headquarters" in Austin, Texas, and has committed acts of infringement in the district.
• Core Dispute: Plaintiff alleges that Defendant’s autonomous vehicle technology, including its Full Self-Driving (FSD) software and associated hardware and training systems, infringes five patents related to predicting human behavior and intent for vehicle navigation.
• Technical Context: The technology at issue involves using supervised machine learning models, trained on human behavioral data, to enable autonomous vehicles to predict the actions and "state of mind" of road users like pedestrians and other drivers.
• Key Procedural History: The complaint alleges that Tesla was aware of the asserted patents prior to this lawsuit, referencing a previous case filed in the Eastern District of Texas. It further alleges that Tesla cited the lead patent (’344) in information disclosure statements during the prosecution of at least 20 of its own patent applications between February and June 2022, which may be relevant to allegations of pre-suit knowledge and willfulness.

Case Timeline

Date	Event
2017-07-05	Priority Date for ’344, ’889, and ’046 Patents
2019-01-30	Priority Date for ’346 Patent
2019-02-06	Priority Date for ’579 Patent
2020-04-07	U.S. Patent No. 10,614,344 Issues
2020-10-01	Tesla releases beta version of FSD software (approx. date)
2021-09-21	U.S. Patent No. 11,126,889 Issues
2022-02-01	Tesla begins citing ’344 patent in its own patent prosecutions (approx. date)
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
2024-01-01	Tesla re-incorporates from Delaware to Texas (approx. date)
2026-03-05	Complaint Filing Date

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

U.S. Patent No. 10,614,344: "System and Method of Predicting Human Interaction with Vehicles" (Issued Apr. 7, 2020)

The Invention Explained

• Problem Addressed: The patent asserts that conventional autonomous driving systems, which rely on predicting "motion vectors" based on past movements, cannot adequately predict the future behavior of people, especially in complex urban environments (’344 Patent, col. 1:35-49).
• The Patented Solution: The invention proposes a system where an autonomous vehicle uses its sensors to capture a road scene, provides that sensor data to a supervised learning-based model, and the model predicts a "statistical summary" of how a plurality of humans would be expected to respond to that scene. The vehicle's operation is then controlled based on this statistical prediction of human intuition, rather than on simple motion extrapolation (’344 Patent, abstract; ’344 Patent, col. 2:9-22).
• Technical Importance: This technology represents a shift from physics-based prediction to a psychometrics-based approach, aiming to imbue an autonomous system with an approximation of human intuition about the intent of other road users (’344 Patent, col. 1:21-31).

Key Claims at a Glance

• The complaint asserts independent claims 1 (method) and 19 (system) (Compl. ¶65).
• Independent Claim 1 includes the core elements of:
  • Receiving sensor data from a vehicle's sensor displaying an object on the road.
  • Providing the sensor data as input to a supervised learning based model.
  • The model is configured to receive the input and predict an "output statistical summary characterizing a distribution of user responses" expected in reaction to the sensor data.
  • Executing the trained model to generate the statistical summary data.
  • Controlling the vehicle's operation based on the generated summary data.
• The complaint reserves the right to assert additional claims, including dependent claims (Compl. ¶65, n.1).

U.S. Patent No. 11,126,889: "Machine Learning Based Prediction of Human Interactions with Autonomous Vehicles" (Issued Sep. 21, 2021)

The Invention Explained

• Problem Addressed: Similar to its parent, the '889 patent addresses the inability of computers to adequately predict human behavior, noting that human drivers have a "natural ability to predict a person’s behavior" that autonomous systems lack (’889 Patent, col. 1:36-41).
• The Patented Solution: The invention details a method for creating and using a model to predict the "state of mind" of road users. The process involves collecting images of road scenes, gathering responses from human observers about the state of mind of users in those scenes, creating a training dataset of summary statistics from those responses, and training a model with it. An autonomous vehicle then uses this trained model to predict the state of mind of a road user in a new, live image and controls its actions accordingly (’889 Patent, abstract; ’889 Patent, col. 2:24-43).
• Technical Importance: This patent focuses on the specific methodology of training a model on human "state of mind" judgments, aiming to move beyond predicting simple actions to predicting underlying intent, awareness, or disposition.

Key Claims at a Glance

• The complaint asserts independent claims 1 (system) and 2 (method) (Compl. ¶85).
• Independent Claim 2 includes the core elements of:
  • Receiving a plurality of images displaying road scenes and a plurality of user responses describing a "state of mind" of a road user in those images.
  • Generating a training dataset comprising summary statistics of the user responses.
  • Training a supervised learning based model with the dataset.
  • An autonomous vehicle receiving a new image of a scene including a road user.
  • The vehicle predicting, using the trained model, summary statistics describing the state of mind of the road user in the new image.
  • Controlling the autonomous vehicle based on the prediction.
• The complaint reserves the right to assert additional claims (Compl. ¶85).

U.S. Patent No. 11,467,579: "Probabilistic Neural Network for Predicting Hidden Context of Traffic Entities for Autonomous Vehicles" (Issued Oct. 11, 2022)

• Technology Synopsis: This patent describes a method for training and using a probabilistic neural network. The network takes an image of traffic as input and generates an output representing "hidden context" (e.g., state of mind, intent) for a traffic entity, along with a "measure of uncertainty" for that output. An autonomous vehicle then navigates based on both the predicted context and its associated uncertainty level.
• Asserted Claims: Independent claims 1 and 20 (Compl. ¶108).
• Accused Features: The complaint alleges Tesla's FSD software is a probabilistic neural network that is trained to generate outputs representing hidden context for traffic entities and navigates the vehicle based on these outputs and their uncertainty (Compl. ¶¶113-120).

U.S. Patent No. 11,520,346: "Navigating Autonomous Vehicles Based on Modulation of a World Model Representing Traffic Entities" (Issued Dec. 6, 2022)

• Technology Synopsis: This patent claims a method where an autonomous vehicle generates a point cloud representation of its surroundings (a "world model"), identifies traffic entities, and uses a machine learning model to determine the "hidden context" of those entities. The system then "modifies" a region of the world model based on this hidden context and navigates the vehicle to stay a threshold distance away from the modified region.
• Asserted Claims: Independent claims 1 and 19 (Compl. ¶131).
• Accused Features: The complaint alleges Tesla's FSD system generates a point cloud representation, identifies traffic entities, determines their hidden context, and modifies its planned path (the "region") to navigate safely (Compl. ¶¶137-144).

U.S. Patent No. 11,753,046: "System and Method of Predicting Human Interaction with Vehicles" (Issued Sep. 12, 2023)

• Technology Synopsis: This patent focuses on the data generation and training process. It describes storing images, sending them to human observers with a request to answer a question about a user's "state of mind," receiving a judgment, generating summary statistics from these judgments, and using this data to train a model. The trained model is then executed by a vehicle to predict a user's state of mind in a new image.
• Asserted Claims: Independent claims 1 and 15 (Compl. ¶155).
• Accused Features: The complaint alleges that Tesla's "Data Engine" and its team of over 1,000 data labelers perform the claimed steps of generating training data from human judgments on road scenes to train the FSD models (Compl. ¶¶159-165).

III. The Accused Instrumentality

• Product Identification: The accused instrumentalities are Tesla's Full Self-Driving (FSD) software (Version 8.0 and subsequent), FSD hardware (HW2 through HW6 and subsequent), Tesla vehicles equipped with FSD (Model S, 3, X, Y, and Cybertruck), and the back-end systems used to train the FSD software (e.g., in-house datacenters, Dojo, and Cortex) (Compl. ¶55).
• Functionality and Market Context: The complaint alleges that the FSD system uses an array of cameras to capture sensor data, which is processed by an onboard computer running neural networks to control the vehicle's steering, acceleration, and braking (Compl. ¶69). A key allegation is that these neural networks are trained using a "Data Engine" that processes vast amounts of video clips collected from Tesla's fleet of vehicles (Compl. ¶¶53; Compl. ¶90). This data is allegedly labeled by an in-house team of over 1,000 people to provide the ground truth for the supervised learning models that predict road user behavior (Compl. ¶70; Compl. ¶91). FSD is positioned by Tesla as a premium driver-assistance feature central to its valuation and future business strategy, including a planned robotaxi service (Compl. ¶17).

IV. Analysis of Infringement Allegations

10,614,344 Patent 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;	The FSD onboard computer receives sensor data from its cameras, which display objects such as other vehicles and pedestrians on the road.	¶69	col. 2:31-35
providing the sensor data as input to a supervised learning based model,	The onboard computer provides the received sensor data as input to the FSD software, which the complaint alleges is a supervised learning based model.	¶70	col. 2:36-37
the supervised learning based model configured to receive an input sensor data ... and predicting 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 user responses associated with the particular object;	The FSD software is allegedly based on supervised learning from Tesla's in-house labelers (the "users") and predicts outcomes that correspond to "a probability mask derived from human demonstration." The complaint's visual evidence shows an "Interaction Search" decision tree that generates "Goal Candidates."	¶70	col. 2:37-43
executing the trained supervised learning based model to generate a statistical summary data characterizing a distribution of user responses expected to be received ...	The FSD software executes the model to generate outputs used for trajectory scoring, which allegedly rely on concepts like "Intervention Likelihood" and a "Human-like Discriminator" built on "Human Demonstrations."	¶70	col. 2:44-49
and controlling the operation of the autonomous vehicle on the road based on the generated statistical summary data.	The FSD software uses the generated output to determine actions for controlling the vehicle, such as deciding whether to "Assert" or "Yield" to a pedestrian.	¶71	col. 2:50-52

• Identified Points of Contention:
  • Scope Question: A primary question may be whether Tesla's internal team of over 1,000 data labelers (Compl. ¶70) qualifies as the "plurality of users" required by the claim. The defense may argue that "users" implies a broader, more diverse set of observers providing subjective responses, as described in some patent embodiments, rather than employees applying structured labels as part of a data processing pipeline.
  • Technical Question: It will be a key factual question whether the output of Tesla's neural network constitutes a "statistical summary characterizing a distribution of user responses." The complaint uses a diagram of an "Interaction Search" tree to support this allegation (Compl. p. 36). The defense may argue that the network's output is a direct navigational plan or cost function, not an intermediate statistical summary of predicted human opinions as contemplated by the patent.

11,126,889 Patent 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 (e.g., Dojo, Cortex) receive millions of video clips from its customer fleet of vehicles. A diagram shows a "Data Engine" that mines the fleet.	¶90	col. 15:43-45
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's team of over 1,000 data labelers assesses the videos and provides advanced labeling that describes a "state of mind," such as selecting from options like "stopped_traffic."	¶91	col. 15:46-49
generating a training dataset comprising summary statistics of uses responses describing the state of minds of road users...;	Tesla allegedly generates a training dataset where the weight of the statistical summary is modified with new responses to improve accuracy over time.	¶92	col. 15:50-53
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 uses the training data set to train the FSD software, which is alleged to be a supervised learning based model that predicts actions based on the possible state of mind of road users.	¶93	col. 15:54-57
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;	A Tesla vehicle in operation receives a new image from its cameras showing a road user, such as a pedestrian on the side of the road.	¶94	col. 15:58-61
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 FSD software in the vehicle makes statistical predictions about the road user's state of mind, evidenced by its prediction of actions like turning or crossing the road.	¶95	col. 15:62-65
and controlling the autonomous vehicle based on the prediction of the supervised learning based model.	The vehicle's operation is controlled based on the prediction, such as by stopping to allow a pedestrian to cross the road.	¶96	col. 15:66-16:2

• Identified Points of Contention:
  • Scope Question: The term "state of mind" will be critical. The complaint alleges that labels like "stopped_traffic" describe a state of mind (Compl. ¶91). The defense may argue that this is merely a label for a physical state, not the cognitive "state of mind" (e.g., intent, awareness) described in the patent.
  • Technical Question: The complaint alleges Tesla generates "summary statistics" of user responses for its training dataset (Compl. ¶92). A technical question is whether Tesla's "Data Engine" (Compl. p. 53) actually aggregates labeler responses into statistical summaries for training, or if it uses the individual labels directly as ground truth for the model, which may not meet the claim limitation.

V. Key Claim Terms for Construction

• The Term: "statistical summary characterizing a distribution of user responses" (from ’344 Patent, claim 1).

• Context and Importance: This term is the inventive core of the ’344 patent, defining the specific nature of the data the model is configured to predict and which is used to control the vehicle. Its construction will likely determine whether the output of Tesla's AI system infringes. Practitioners may focus on this term because modern neural networks often produce probabilistic outputs, and the dispute will center on whether those outputs meet this specific definition.

• Intrinsic Evidence for Interpretation:

  • Evidence for a Broader Interpretation: The patent's abstract states the device "aggregates a subset of the plurality of response data to form statistical data and a model is created based on the statistical data." This could support an interpretation where any model trained on aggregated human responses and producing a probabilistic output meets the limitation.
  • Evidence for a Narrower Interpretation: The specification explicitly lists examples of what the statistical data may be associated with, including "at least one of a central tendency, a variance, a skew, a kurtosis, a scale, and a histogram" (’344 Patent, col. 2:44-46). A defendant may argue that the "statistical summary" must comprise one or more of these specific statistical metrics, rather than just a general probabilistic output from a neural network.

• The Term: "state of mind" (from ’889 Patent, claim 2).

• Context and Importance: This term distinguishes the '889 patent's claims from simple action prediction. The infringement analysis will depend on whether the information Tesla's labelers provide, and which the FSD system subsequently predicts, legally constitutes a "state of mind."

• Intrinsic Evidence for Interpretation:

  • Evidence for a Broader Interpretation: The patent claims user responses "describing a state of mind," which could be argued to encompass any label that requires a judgment beyond simple object identification, including classifying a vehicle's activity (e.g., "stopped_traffic").
  • Evidence for a Narrower Interpretation: The specification of the parent '344 patent, incorporated by reference, gives examples of what a "state of mind" judgment may entail, such as "intention, awareness, personality, state of consciousness, level of tiredness, aggressiveness, enthusiasm, thoughtfulness" (’344 Patent, col. 9:36-40). A defendant could argue that the predicted "state of mind" must relate to these specific types of cognitive or psychological attributes, not merely a classification of a physical situation.

VI. Other Allegations

• Indirect Infringement: The complaint alleges that Tesla induces infringement by providing customers with FSD-equipped vehicles and instructing them on its use through manuals, advertisements, and test drives (Compl. ¶¶58-59). The complaint specifically notes that Elon Musk mandated FSD demonstrations for every prospective buyer in North America, which it alleges is an act of encouraging and instructing customers to use the infringing functionality (Compl. ¶59).
• Willful Infringement: The complaint alleges willful infringement based on Tesla’s alleged pre-suit knowledge of the patents (Compl. ¶62). This allegation is supported by two key facts: (1) the existence of a prior patent infringement lawsuit between the parties in another district (Compl. ¶29), and (2) Tesla’s alleged citation of the ’344 patent as relevant prior art during the prosecution of its own patent applications at the USPTO, starting between February and June 2022 (Compl. ¶30).

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

• A core issue will be one of definitional scope: can claim terms like "statistical summary characterizing a distribution of user responses" and "state of mind" be construed to cover the outputs and training labels of a modern, end-to-end autonomous driving neural network? The case may turn on whether Tesla's system, which is trained to produce a safe driving path, can be legally characterized as predicting a statistical distribution of human opinions about road user intent.
• A central evidentiary question will be one of technical implementation: what is the precise nature of the data that Tesla’s "Data Engine" provides to its neural network for training, and what is the precise nature of the data the in-vehicle FSD computer generates as an intermediate step before making a control decision? The infringement analysis will depend on whether discovery reveals a direct mapping between the internal workings of Tesla's FSD system and the specific processes recited in the claims.
• A key question for damages will be willfulness: does Plaintiff's allegation that Tesla cited the '344 patent as prior art during its own patent prosecution establish pre-suit knowledge of the patent and its relevance, thereby exposing Tesla to the risk of enhanced damages for any infringement found after that date?