Robotic goalie with 3 ms reaction time at 4% CPU load using event-based dynamic vision sensor

Area of Science:

• Robotics
• Computer Vision
• Neuromorphic Engineering

Background:

• Conventional vision systems face high computational costs and latency.
• Robotic systems require fast response times for dynamic tasks.

Purpose of the Study:

• To develop a fast, self-calibrating robotic goalie using a dynamic vision sensor (DVS).
• To achieve high update rates and low latency at reduced CPU load.

Main Methods:

• Utilized an asynchronous neuromorphic dynamic vision sensor (DVS) silicon retina.
• Implemented per-pixel event-based tracking of multiple balls based on illumination changes.
• Developed self-calibration for the robotic arm's motor output map.
• Planned open-loop arm movements for precise ball interception.

Main Results:

• Achieved median update rates of 550 Hz with latencies of 2.2 ± 2 ms.
• Demonstrated a blocking capability of approximately 80% for fast shots from 1 meter.
• Maintained peak CPU load below 4% during operation.
• Provided practical measurements of USB device latency.

Conclusions:

• The DVS enables a low-latency, high-update-rate robotic goalie system.
• The system offers efficient performance with significantly reduced computational requirements.
• The DVS-based approach is effective for fast-paced robotic applications.