Skip to content
Case Study

VEX AI 16610A Level 5 — inaugural Robot Intelligence Award

In 2026, VEX AI Team 16610A Level 5 received the inaugural Robot Intelligence Award at a VEX AI event held at UC Berkeley — recognition not for a single model or routine, but for a complete robotic intelligence system operating in real time.

Context

VEX AI competitions require robots to operate autonomously in a dynamic environment: moving game elements, an opposing alliance, and strict time budgets. 16610A Level 5 competed at the highest autonomy level at a VEX AI event held at UC Berkeley in 2026.

Challenge

The robot must continuously read its environment, estimate its own state, understand task changes, evaluate multiple possible actions, and execute stable control — under uncertainty, on limited onboard compute, with no human in the loop.

Constraints

  • All inference on the robot's onboard edge device — no cloud round-trips
  • Real-time budgets: perception, planning and control inside one loop
  • Dynamic environment: game objects and opponents move constantly
  • Competition rules: no human input during autonomous operation

System Design

  • Perception: edge vision inference identifying targets, obstacles and task state
  • Localization: multi-sensor fusion for continuous position, orientation and motion estimates
  • Decision-making: RUR1's proprietary patent-grade strategic planning and decision-making algorithm, weighing expected value, execution time, travel cost, resource state, conflict probability and operational risk
  • Execution: dynamic path and trajectory generation with closed-loop motion control
  • Safety: anomaly handling and fallback behaviors when conditions degrade

Implementation

The stack runs as a real-time loop on the robot. Perception feeds a fused world state; the strategic planning layer evaluates candidate actions against the live game state within tight time limits; selected strategies become paths, trajectories and motor commands, corrected continuously through execution feedback. The robot adjusts task priorities and re-evaluates its plan as field conditions change.

Measured Results

  • Autonomous navigation and strategic decision-making in live matches

    Source: 16610A codebase and system architecture documentation · 2026-05-01

  • Inaugural Robot Intelligence Award — VEX AI, UC Berkeley, 2026

    Source: VEX AI event results, UC Berkeley, 2026 · 2026-05-01

What We Learned

  • Meaningful autonomy emerges from the reliable coordination of perception, understanding, decision-making, planning and action — not from any single algorithm.
  • Multi-sensor fusion beats any single sensor: each source fails differently, and the fusion covers the gaps.
  • Strategic planning needs safe fallbacks: when conditions degrade, the robot must stay productive, not freeze.
  • Testing under competition pressure finds failure modes no lab run reproduces.

Technology

  • Environmental perception
  • Edge AI inference
  • Multi-sensor localization
  • State estimation
  • Patent-grade strategic planning
  • Dynamic path & trajectory generation
  • Closed-loop motion control
  • Safety & fallback mechanisms
  • Full-stack system integration