The Rover Cam rolls on large wheels that handle small obstacles and uneven terrain, letting it move through spaces where fixed cameras and simpler wheeled bases would struggle to follow.

Preorders open at $999.99 with shipping expected to start December 16, positioning the robot as an affordable option for businesses wanting proactive security instead of passive monitoring.

The telescoping periscope camera records 4K video and can raise or lower to set heights during scheduled patrols, a bigger role than it played on the earlier Astro hardware.

Battery life is estimated at about 30 minutes, and coverage is limited to roughly 5,000 square feet on one floor, so multi-level sites need extra robots or a Ring Drone Cam.

Amazon previously launched Astro for Business in 2023 and discontinued it just one year later, making the Ring Rover Cam a second attempt at business robotics under the new Ring Robotics brand.

The Ring Drone Cam weighs 165 grams and can fly upstairs to monitor a room, with spotlights that activate in dark environments to make the home look occupied.

Flight time is five to eight minutes per charge, which is enough for a scheduled patrol but not for long unattended surveillance sessions across a large house.

The drone runs a motor health check before every take-off and refuses to fly if it detects a problem, lands on its own if internet connectivity is lost.

Video is encrypted in transit using TAKE, or Throw Away the Key Encryption, and the propellers are intentionally loud to make remote snooping harder for bad actors.

The Ring Rover Cam covers up to 5,000 square feet with a telescoping 4K camera, two-way talk, and automatic return to its charging dock after each patrol ends.

The home security drone flies room to room, shoots 2K video, and uses LED lights for night and low-light recording, retailing at $270 and shipping January 27.

The drone can be scheduled to patrol at specific times and controlled through Amazon's Alexa Home Map, tying it into the broader Ring and Alexa smart home stack.

The enterprise Rover streams continuous 4K video, includes a periscope camera and two-way audio, and automatically records when it detects a human presence in the area.

Both robots use AI Pilot, Ring's proprietary system that combines specialized vision models, sensors, and cloud processing to navigate autonomously and build schematics of spaces.

Ring says the Rover can sweep floors and investigate sensor trips at 2 a.m., with the $1,000 unit shipping December 16 and encrypted under TAKE, Ring's strongest default encryption yet.

The K3 Scout is an 8.4 m uncrewed boat capable of 55 knots, carrying payloads up to 600 kg across ranges of 650 nautical miles.

The Rattler is a 7.2 m boat with a 500 kg modular bay and up to 950 nautical miles of range, built for longer missions at sea.

The squadron has around 180 personnel split into two watches, giving the Royal Navy round-the-clock operations for its autonomous fleet.

Last year, five uncrewed boats were remotely piloted to escort HMS Tyne and mine countermeasures mothership HMS Stirling Castle in a demonstration.

The squadron replaces Project Beehive, the navy's experimental program, and gets 20 boats from Kraken Technology Group as its operational core.

Holoscan 5 groups operator chains into partitions on a single thread, cutting cross-thread handoffs and unnecessary syscalls that inflate control latency and jitter in tight loops.

Clock and data-arrival triggers schedule work, while bounded drop-oldest queues stop stale sensor frames from piling up behind fresh commands during a control cycle.

The Seeed Studio reBot Arm B601-DM uses quasi-direct-drive actuators, giving the tutorial a low-cost arm with high torque bandwidth for both control and sensing work.

Pinocchio loads the arm's URDF model and runs a Recursive Newton-Euler Algorithm solver, computing inverse dynamics from link masses, dimensions, and moments of inertia.

The tutorial requires Python 3.12 or newer on Linux, and Nsight Systems annotations built into the Holoscan SDK let you measure latency and jitter per operator.

Gemini 4 Argon scores 87% on the Embodied Reasoning Question Answering (ERQA) benchmark using pure reasoning, without external tools or robot feedback in the evaluation loop.

A vision-language-action policy fine-tuned on domain-randomized MuJoCo episodes synthesized from one RGB photo rose from 56% to 75% success on the corresponding real-world setup.

Google's Real2Sim examples rebuild MuJoCo scenes from teleoperation video, and the simulation tuning matches contact events as well as visuals, not just pixel-level appearance.

The model is available to Google AI Ultra subscribers and through the paid Gemini API, with model ID gemini-4-argon listed in the official Gemini API models documentation.

Google's Gemini 4 Argon for Robotics dev blog documents the full Video-to-Sim-to-Real pipeline, but it publishes no per-demo failure rates beyond that single vision-language-action case.