Deep Robotics Lite3 Basic AI

The Lite3 Basic AI walks, trots, and runs on flat and uneven terrain, climbs stairs up to about 18 cm (7.1 inches), recovers from being pushed or tipped, and follows preprogrammed gait patterns. It has a built-in IMU and joint encoders so it can balance and self-stabilize without external sensors. Out of the box you can drive it manually with the included wireless controller and run the demo behaviors. Anything beyond that requires writing code against the SDK.

The Pro adds an onboard NVIDIA Jetson compute module, factory-installed depth cameras, and a higher-capacity battery. The Basic AI comes with the same chassis, motors, and locomotion controller but no perception sensors and no onboard AI computer, so you supply your own if you want to add vision, mapping, or autonomy. Pick the Basic AI if you plan to add custom perception hardware. Pick Pro if you want to start running navigation algorithms immediately.

The Lite3 ships with a C++ SDK and a Python SDK. ROS 1 (Noetic) and ROS 2 (Foxy) packages are published on Deep Robotics' GitHub. You can send walk, trot, stand, and pose commands at the locomotion layer or write your own controllers at the joint-torque level. URDF files for Gazebo and MuJoCo simulation are included so you can develop without hardware.

Continuous walking runtime is approximately 1.5 to 2 hours per charge. Standby time is up to 4 hours. The battery is removable and hot-swappable, so a second pack lets you keep working without downtime. Charging from empty takes about 40 minutes to 1 hour with the included AC adapter. Lithium-polymer cells should not be discharged below 10 percent or stored fully charged for long periods.

Yes within reason. The Lite3 handles short grass, packed gravel, dirt trails, and dry leaves without issue. Loose sand, mud, and accumulated snow over a few centimeters can compromise traction and may require slower gait settings. DeepRobotics does not assign the Lite3 an official IP rating, and the manufacturer's manual advises against operating it in rain, snow, or wet conditions, so treat it as a dry-environment platform. Always wipe away any incidental moisture and avoid standing water, since moisture intrusion can cause damage.

Rated continuous walking load is 5 kg (11 lb), up to a 7.5 kg (17 lb) maximum, distributed across the back deck. The deck has M5 threaded mounting points at standard spacing for cameras, lidar, robotic arms, and battery extenders. Heavy or off-center loads reduce maximum walking speed and stair-climbing capability. Mount sensors as low and centered as possible to keep the center of gravity within the support polygon.

Yes. Deep Robotics provides URDF and MJCF files compatible with NVIDIA Isaac Sim, Gazebo Classic, Gazebo Sim (Ignition), MuJoCo, and PyBullet. The locomotion controller has a simulation mode that mirrors the real robot's joint behavior, so policies trained in sim can be deployed to hardware with minimal sim-to-real gap. Sample reinforcement learning environments are included in the SDK repository.

The Lite3 Basic AI comes with the robot, one battery, an AC charger, a wireless controller, and a USB-to-Wi-Fi development link. To program it you need a Linux laptop running Ubuntu 20.04 with ROS installed, plus a network connection to the robot. No external power supply is needed during operation since it runs entirely on the onboard battery.

The Lite3 Basic AI carries a 12-month manufacturer warranty covering motors, electronics, and the locomotion controller. Wear items like rubber foot pads and battery cells are not covered after 6 months. Warranty service is handled through RoboticsSelect as the authorized dealer. Email support@roboticsselect.com or call (412) 415-7375 for a return authorization. International shipping for repairs is at the customer's expense.

Yes. Each unit has independent compute and locomotion control, so you can run multiple robots in the same area as long as each one is running its own SDK instance. For coordinated swarm behavior you need to implement the higher-level coordination yourself, typically through ROS multi-master or a centralized planner. The robots do not share state automatically.

Sound output during a normal trot is approximately 65 to 70 dB at 1 meter (3.3 feet), comparable to a vacuum cleaner. Most of the noise comes from the harmonic drive gearboxes in the joints. Standing still or in low-power standby it is nearly silent. Indoor use is fine for offices and labs but the noise can be intrusive in classroom recording or video shoots.