Deep Robotics Lite3 Venture

The Lite3 Venture runs locomotion gaits trained through reinforcement learning, which improve stability and terrain adaptability compared to the standard Basic's conventional gait controller. The hardware is otherwise identical to the Lite3 Basic: it does not add an onboard computer, depth cameras, or other perception sensors. Choose the Venture if you want the improved reinforcement-learning locomotion out of the box.

No. The Venture ships with the same sensing as the Basic: a built-in IMU and joint encoders for balance, plus the wide-angle camera and ultrasonic radar of the base platform. It does not include depth cameras, lidar, or an onboard AI computer. You can add your own perception hardware through the chassis USB and Ethernet interfaces. Common additions include the Intel RealSense D435i for depth, a Livox lidar, and StereoLabs ZED cameras for stereo vision.

The Venture does not include an onboard AI computer, so you would add your own compute module, for example an NVIDIA Jetson, connected through the chassis USB and Ethernet interfaces. The locomotion controller exposes a standard interface, so an external computer running your vision or policy code can publish velocity commands while the robot handles balance and gait. The open C++ and Python SDK supports this integration.

The Pro adds hardware the Venture does not have: an onboard NVIDIA Jetson Xavier NX perception computer, a factory-installed Intel RealSense D435i depth camera, and additional hardware interfaces. The Venture is the Basic platform with reinforcement-learning-trained gaits and no added perception hardware. Choose the Venture for improved locomotion at a lower cost; choose the Pro if you want onboard perception and compute out of the box.

You develop on a Linux laptop running Ubuntu 20.04 and communicate with the robot over Wi-Fi or Ethernet using the Deep Robotics SDK. The locomotion controller is exposed through the SDK and ROS, so your nodes send velocity and gait commands while the locomotion layer handles balance. ROS 1 (Noetic) and ROS 2 (Foxy) packages are published on Deep Robotics' GitHub.

Both are achievable but require you to add perception hardware and compute, since the Venture does not include cameras, lidar, or an onboard AI computer. Person-following typically uses a depth camera and a detection model on an added compute module; SLAM-based navigation requires a lidar and a navigation stack such as Nav2. Deep Robotics provides example code, but these are not turn-key features on the Venture.

Continuous walking runtime is approximately 1.5 to 2 hours per charge, the same as the standard Basic, since the Venture uses the same 28.8 V, 4.4 Ah hot-swappable battery. Charging from empty takes about 40 minutes to 1 hour. If you add your own compute and sensors, their power draw will reduce runtime accordingly. A second battery enables continuous operation through hot-swapping.

The Venture's locomotion policies are trained through reinforcement learning, which improves balance and adaptability on slopes, stairs, and irregular terrain compared to the standard Basic's controller. The mechanical platform, motors, and 12 degrees of freedom are identical between the two. The difference is in the control software, not added hardware.

Yes. The Lite3 platform appears in published research on locomotion learning, multi-robot coordination, and vision-language navigation. The open SDK and standard hardware make experiments reproducible by other labs. Cite Deep Robotics' platform documentation and include URDF version numbers in your methods sections to support reproducibility.

12-month manufacturer full-unit warranty, with core components such as the battery and joints covered for 6 months. RoboticsSelect handles all warranty claims as the authorized dealer. Wear items including rubber foot pads, battery packs after 6 months, and cosmetic damage from drops are not covered.