PTAB

IPR2023-01054

Tesla Inc v. Autonomous Devices LLC

Log In
Key Events
Petition
petition

1. Case Identification

2. Patent Overview

  • Title: Learning Device Operation in Circumstances for Autonomous Operation
  • Brief Description: The ’974 patent discloses systems and methods for autonomous device operation. The technology involves a device learning from a user's operation by correlating "circumstance representations" (e.g., sensor data like images) with corresponding instruction sets, storing these correlations in a knowledgebase, and later using this knowledgebase to autonomously execute instructions when a new, matching circumstance is detected.

3. Grounds for Unpatentability

Ground 1: Anticipation of Claims 1-9 and 12-20 by Grotmol

  • Prior Art Relied Upon: Grotmol (Patent 9,604,359)
  • Core Argument for this Ground:
    • Prior Art Mapping: Petitioner argued that Grotmol anticipates every limitation of the challenged claims. Grotmol taught a robotic device that learns associations between sensory context (images) and respective actions (motor commands) during a user-led training process. This directly maps to the ’974 patent’s core method, including:
      • Accessing a memory storing a knowledgebase of correlations (Grotmol’s "training buffer" storing images associated with motor commands).
      • Learning these correlations through a process involving user operation (Grotmol’s "listen mode" or "override mode" where a trainer operates the robot).
      • Receiving a new circumstance representation (a current camera image).
      • Anticipating an instruction set based on a "partial match" by comparing the new image to stored images to find the "best matching stored image."
      • Executing the correlated instruction set to operate the device autonomously.
    • Key Aspects: Petitioner contended that Grotmol’s disclosure of loading a "trained configuration" to "one or more other robots" anticipates the claims’ recitation of operations involving a "second device."

Ground 2: Obviousness of Claims 10 and 11 over Grotmol and POSITA Knowledge

  • Prior Art Relied Upon: Grotmol (Patent 9,604,359) in view of the knowledge of a Person of Ordinary Skill in the Art (POSITA).
  • Core Argument for this Ground:
    • Prior Art Mapping: Claims 10 and 11 add the limitation that an additional correlation is learned during a process involving operation of the device by another user. Petitioner asserted that Grotmol taught training multiple behaviors, including both "simple tasks" and "complex tasks."
    • Motivation to Combine: A POSITA would have found it obvious to use different trainers (users) to train different behaviors. This would allow a system to take advantage of the varying skills and knowledge of different trainers, such as having an expert train a complex task while a novice trains a simple one. Grotmol did not limit the number of users who could train the robot.
    • Expectation of Success: The application of this common-sense approach to Grotmol’s system would have been straightforward and yielded predictable results, specifically an expanded knowledgebase with correlations from multiple users.

Ground 3: Obviousness of Claims 4 and 11 over Grotmol in view of Hickman

  • Prior Art Relied Upon: Grotmol (Patent 9,604,359) and Hickman (Patent 8,639,644).
  • Core Argument for this Ground:
    • Prior Art Mapping: This ground addressed limitations requiring the knowledgebase to include correlations learned from the operation of a second device. Grotmol taught that a "trained configuration" could be saved and loaded onto "other robots to provide learned behaviors." Hickman taught a "shared robot knowledge base" using cloud computing where learned experiences from one robot are uploaded and made available to other robots, allowing them to "benefit from the prior experiences of other robots."
    • Motivation to Combine: A POSITA would combine these references to improve the efficiency and scale of sharing learned behaviors as taught by Grotmol. Hickman provided an express, advantageous mechanism (a cloud-based shared knowledgebase) for implementing Grotmol’s concept of sharing trained configurations among multiple robots. Petitioner noted that Hickman was cited on the face of Grotmol and that both references address the same technical field and similar problems, such as learning a "fetch" task.
    • Expectation of Success: Combining Grotmol’s local training system with Hickman’s cloud-based sharing architecture was presented as a predictable integration of complementary technologies. The combined system would predictably result in a more robust and scalable learning platform where training sets from multiple devices could be aggregated and shared, as claimed.

4. Key Claim Construction Positions

  • "means for generating or receiving," "means for anticipating," "means for executing" (Claim 19): Petitioner argued that these means-plus-function limitations require construction under 35 U.S.C. §112, ¶ 6. Based on the patent’s specification, Petitioner asserted that the corresponding structure for performing all three of these functions is "one or more processor circuits." This construction was central to its argument that the system claims (19-20) are anticipated for the same reasons as the corresponding method claims.

5. Relief Requested

  • Petitioner requests institution of an inter partes review and cancellation of claims 1-20 of the ’974 patent as unpatentable.