We’re bringing Matrix, our outdoor security robot, to ROSCon 2025, and we’re proud to collaborate with NVIDIA through the NVIDIA Inception program as we continue advancing the future of autonomous security robotics.
Matrix operates in complex outdoor environments such as campuses, parking lots, and industrial parks, where on-premise processing is critical. Security operations often require analytics to remain local for privacy, latency, and reliability reasons, making edge AI a core part of our deployment strategy.
To meet these needs, Matrix features the industrial NVIDIA Jetson AGX Orin acquired from Advantech as one of its onboard computers. Jetson Orin provides the GPU performance and power efficiency needed for real-time video analytics, allowing Matrix to detect and track people, vehicles, and security-related events such as suspicious behavior, loitering, and unattended bags, all processed locally without cloud dependence.
Matrix integrates five dedicated security camera streams, each feeding into a GPU-accelerated video analytics pipeline. The platform uses NVIDIA NVDEC hardware decoders to read frames from RTSP streams and NVENC hardware encoders to output annotated video with bounding boxes and tracking overlays. These dedicated video blocks offload compression and decompression from the GPU so it can focus on AI inference and multi-object tracking, which is essential for maintaining high-throughput, real-time analysis at the edge. The processed video streams and analytics results are streamed directly to the Security Control Center (SCC), where security officers can monitor detections and tracking data.
We use NVIDIA Triton Inference Server to manage and run multiple analytics models in parallel, supporting multicamera detections, multi-object tracking, and robot-on-the-move analytics that maintain continuous situational awareness during patrols. ROS 2 Jazzy provides the backbone for robot control, communication, and coordination across the system.
Beyond deployment, NVIDIA accelerated computing powers our training and simulation workflows, enabling rapid iteration and validation in Gazebo before field rollout.
Looking ahead, we’re expanding how we leverage NVIDIA technologies across development and autonomy. With the NVIDIA Isaac open robotics development platform, we plan to simulate large outdoor environments, auto-generate annotations and runnable paths, and improve forward detection and prediction of dynamic objects for safer, more intelligent navigation. We’re also integrating NVIDIA Video Search and Summarization (VSS) Blueprint with NVIDIA Cosmos Reason VLM to develop agentic security analytics, enabling robots to interpret video captured by the robot, summarize security events, and generate contextual insights autonomously.
By combining NVIDIA Jetson, Triton, and ROS 2, we’re building reliable, on-premise robotic systems capable of delivering real-time intelligence directly from the field. If you’re attending ROSCon 2025, visit our booth to see Matrix in action and learn how our collaboration with NVIDIA is driving the next generation of security robotics.