World’s First Autonomous BJJ Humanoid
Executive Summary
By Dorian Cartwright | TrustRobotics™
Detailed Architecture Guide
Explore each pillar through its own focused detail page. This guide connects the topics and explains tap handling, the continuous control loop, and development targets.
I. Real World Perception: Sensor Fusion for BJJ Pose Estimation
Video, tactile contacts, joint encoders, and inertial measurement units (IMUs) are fused to estimate body position, orientation, and motion. The perception state also considers center of mass, inertia, momentum, detected resistance, and grip position and strength, distinguishing direct measurements from model-based estimates and retaining uncertainty.
II. Real World Action: Inter-Body Artificial Intelligence
Inter-body artificial intelligence interprets grips, frames, support, loading, and constraints to evaluate available actions. The BJJ LLM trained in section III uses this relational and physical state, training objectives, and interaction history to select an intended response and output a structured BJJ action token. Tokens form an action timeline for downstream validation and motion control; the physical exchange determines the actual next state.
III. Teaching a robot a jiu-jitsu lineage
Section III trains and adapts the BJJ LLM to an instructor’s jiu-jitsu persona. Demonstrations, corrections, and supervised exchanges connect physical and relational states with preferred actions, positional priorities, teaching objectives, and responses to resistance. Section II uses the resulting model during the roll to output BJJ action tokens.
IV. A robot must also know when to tap
A credible BJJ robot must understand submission in both directions.
The system can detect a participant’s physical tap, verbal submission, release condition, excessive force, joint condition, or other biomechanical safety condition. In response, the robot can limit force, stop or reverse an action, release a grip, or transition to a safer state.
The robot can also determine when it should submit. It does not need to wait until a joint reaches a mechanical failure threshold. If the relational state indicates that the robot has been caught in a completed submission, it can tap, cease resistance, or disengage.
Intelligence on the mat includes knowing when the match is over.
Training objectives and execution constraints
Outcome prediction and instructional preference are distinct. The most likely successful action may not be the most useful teaching action. A teaching policy may preserve a position to let a student recognize and practice a response. All such preferences operate within execution constraints.
Termination preempts the timeline
A tap, verbal submission, or other stop condition interrupts the remaining action-token timeline. The robot performs an appropriate controlled release or disengagement and checks the resulting state. A favorable completion probability never overrides a termination condition.
Related patent
DOR-282 — Brazilian Jiu-Jitsu Humanoid Robot Physical Action Control. This overview includes proposed implementation details; the filed application defines its own disclosure and claim scope.
← AUTONOMOUS BJJ HUMANOID OVERVIEW
Proposed Architecture
Continuous perception-to-action loop
OBSERVE: collect video, tactile, joint, and other available measurements.
ESTIMATE: maintain both body poses, contact relationships, and uncertainty.
INTERPRET: identify the current BJJ state and available transitions.
SELECT: evaluate candidate outcomes and choose an intended transition.
EXECUTE: generate an action-token timeline for constrained motion control.
REASSESS: update, continue, interrupt, or replace the plan using the resulting interaction.
Technical Specifications
Technical development targets
Multimodal perception: time-aligned observations with confidence.
Pose and contact estimation: body configurations and associated contact estimates.
Relational-state model: physical relationships and BJJ state hypotheses.
Transition evaluator: conditional outcome estimates for candidate actions.
Action-token generator: time-ordered objectives, constraints, and interruption conditions.
Execution and termination control: feedback-driven movement, stop handling, and release confirmation.
These are development targets, not published hardware specifications or measured performance claims.
Partner with
TrustRobotics
Integrate the industry's premier
AI-to-actuator safety layer.
Accelerate Open Development
Validate & Test Stack
Secure Commercial Licensing