LSR100 Indoor Differential Service Robot Platform

Key Features & Open-Source Development Resources

The LSR100 development library includes the following functions. Some AI, voice, outdoor-navigation, multi-line LiDAR, cellular communication and autonomous-charging functions require the corresponding optional hardware or compute configuration.

01 AI Large Model Integration (Requires Voice Module)

• Large Language Models (LLM): Supports real-time access to major AI model platforms, enabling deep semantic parsing of text commands and flexible, intelligent conversational responses.
• Large Voice Models: Powered by AI voice models, featuring real-time speech-to-text and text-to-speech conversion. Combined with on-board voice modules and speakers, intelligent interaction is more seamless than ever.

• Large Vision Models: Leverages high-definition cameras to rapidly analyze and comprehend image content, precisely identify object categories, and provide both text and voice feedback.

• Local Model Deployment: Supports local deployment of large models for low-latency AI inference. By eliminating cloud dependency, the autonomous decision-making capabilities of the intelligent device are significantly enhanced.

02 AI Multi-modal Large Model Applications (Requires Voice Module)

• AI Vision: Scene Understanding & Perception: Analyzes images/videos via vision large models to identify environmental elements and relationships, providing cognitive feedback.

• AI Recognition: Object Tracking: Locks onto targets through vision model analysis to perform real-time dynamic tracking.

• AI Interaction: Color-Aware Line Following: Uses Large Language Models (LLM) to identify color information for interaction, precisely sensing lines for path following.

• Flexible Chassis Control: AI-empowered chassis control with LiDAR-based emergency stop capabilities for safe and intelligent maneuvering.

• AI Interaction: LiDAR Tracking: LLM-activated LiDAR tracking for seamless transitions between convenient control and precision tracking.

• Built-in RAG + Dual-Model Inference Architecture: Retrieves semantic fragments from knowledge bases; LLMs handle task decomposition and step planning, while multi-modal models convert instructions into executable Vision-Action closed-loop commands.

• Voice Wake-up & Interruption Support: Allows voice commands to interrupt ongoing operations to receive and process new dialogue instructions.

03 Embodied AI SLAM Mapping & Semantic Navigation

• AI Multi-modal Large Model + Semantic Control Mapping: Understands semantics through AI multi-modal large models. Once mapping is activated, the robot is controlled via semantic commands to scan and save maps of the area.

• AI Multi-modal Large Model + Semantic Navigation: Set semantic coordinate files within the pre-built map. During navigation, these are loaded automatically; the robot understands semantic meanings and assigns navigation target points accordingly.

• AI Multi-modal Large Model + Voice Semantic Labeling: Add new semantic coordinate labels via voice dialogue during navigation without pre-writing configuration files. Low environmental requirements with high labeling flexibility.

• AI Multi-modal Large Model + Full-Function Combined Application: Leveraging AI multi-modal voice interaction, all robot functions can be combined without relying on specific environment setups. Supports auto-start of interactive nodes, allowing users to master ROS functions with zero coding for a quick start.

04 Outdoor Navigation & LiDAR Perception ( Optional)

Multi-Line LiDAR Integration: Designed for low-speed autonomous vehicles, our 16-line mechanical LiDAR utilizes TOF (Time-of-Flight) ranging.

Note: The Multi-Line LiDAR comes standard on the Ultra version.

Autoware Outdoor Mapping & Navigation

Complex Path Planning: Achieve advanced navigation using generated High-Definition (HD) maps.

Point Cloud Obstacle Avoidance: Real-time navigation safety powered by LiDAR point cloud clustering.

Recommended Configuration: For optimal Autoware performance, we recommend the NVIDIA Orin-series controller paired with 16/32-line LiDAR.

05 Autonomous Power Management ( Optional)

06 Brand-New Premium Hub Motors

• Powered by brushless motors and equipped with a pendulum suspension system and silent caster wheels, ensuring ultra-smooth and noise-free operation for the entire robot.

Fully Open-Source ROS Feature Library

1.RTAB-Map 3D VSLAM & LiDAR Integration

Supports RTAB-Map pure vision mapping and LiDAR-vision fu sion mapping. Fully compatible with ROS 1 and ROS 2 for versatile 3D environment reconstruction.

2.Classic 2D LiDAR Mapping & Navigation

ROS 2: Supports Gmapping, Cartographer, and slam_toolbox.

ROS 1: Supports Gmapping, Hector, Karto, and Cartographer.

Navigation: Features autonomous point-to-point navigation, multi-point waypoints, and dynamic obstacle avoidance.

3.ORB-SLAM2 Visual Mapping

Features the open-source ORB-SLAM2 framework for real-time camera pose

estimation and sparse 3D reconstruction. Provides real-scale metric information when used in RGB-D mode.

4.ROS QT Graphical User Interface (GUI)

Deployed with a dedicated QT-based GUI for "one-click" ROS activation. Provides

intuitive real-time feedback on robot velocity, battery status, and system health.

5.YOLO Object Detection

ROS 1: Powered by YOLO.v3 for general object, traffic sign, and gesture recognition.

ROS 2: Powered by YOLO.v8 and YOLOv11 for state-of-the-art object detection and custom model training support.

6.LLM Deployment

Supports both offline local deployment and online cloud-based integration of Large

Language Models for advanced reasoning and interaction.

7.Depth-Based Visual Following

Utilizes depth cameras to calculate target distance and bearing for smooth, real-time robot following.

8.KCF Target Tracking

Employs Kernelized Correlation Filters (KCF) via the depth camera to identify and track objects with fixed visual features.

9.AR Tag Recognition & Following

Detects and tracks the 6-DOF pose of AR Tags, allowing for tag-following behaviors and expanded tag-based localization.

10.RRT Autonomous Exploration

Enables fully autonomous mapping using the RRT algorithm. The robot explores, maps, saves the data, and returns to the starting point without human intervention.

11.Web-Based Camera Monitoring

Remotely view the robot's live camera feed through any PC browser for quick deployment of remote surveillance tasks.

12.RGB Camera Line Following

Enables the robot to follow ground lines via RGB vision. Integrated with LiDAR to ensure automatic obstacle avoidance during line-following missions.

13.LiDAR-Based Following

Scans the environment for nearby obstacles and intelligently selects the nearest target for the robot to follow.

14.LiDAR Angle Masking

Optimized via SDK to allow custom angle shielding/masking for all supported LiDAR models.

15.Sound Source Localization

Uses a microphone array to achieve sound source localization with 1° precision. This technology facilitates advanced noise reduction and voice-driven navigation.

16.Voice-Activated Summoning

Call the robot to your location from anywhere. It identifies the user's direction via the mic array and determines distance using LiDAR.

17.Voice-Controlled Navigation

Semantic analysis of voice commands allows the robot to navigate to any predefined point on the saved map.

18.TTS (Text-to-Speech) Interaction

Enables full human-machine interaction through TTS technology, with support for expanded iFLYTEK online voice dialogue features.

19.TEB & DWA Path Planning

Includes detailed video tutorials and Python-based "mini-games" to help users learn navigation path planning from the ground up.

20.DWB / MPPI / RPP Controllers

Provides three ready-to-use controller plugins (DWB, MPPI, and RPP) tailored for different scenarios and various robot footprints.

21.Full Coverage Path Planning

Automatically generates a path that covers an entire user-specified area, ideal for cleaning or inspection robots.

22.Chassis Kinematics Analysis

Provides comprehensive kinematics analysis for various chassis types: Ackermann, Differential, Tracked, Mecanum , Omnidirectional, and 4WD.

23.Path Recording & Playback

Record manual trajectories and reproduce them autonomously using the Nav2 navigation framework.

24.Comprehensive URDF Models

Includes high-fidelity URDF models that accurately match the physical robot's dimensions and properties.

25.ROS Mobile APP for Navigation

A dedicated app for controlling the ROS environment, supporting motion control, mapping, and navigation tasks.

26.Hardware Tuning & Parameter APP

Supports Android and iOS. Features real-time parameter tuning, gravity-sensing control, and waveform visualization.

27.Wireless Code Flashing (Bluetooth)

Standard Bluetooth module allows for remote debugging and "second-level" code flashing via the dedicated host computer, simplifying secondary development.

28.Optional 4G/5G Remote Modules

Supports optional 4G/5G modules for long-range video transmission and remote control over cellular networks.

29.3D Reconstruction for Real-World Autonomous Driving

Upgrade with an optional multi-line LiDAR to unlock outdoor 3D mapping and environment reconstruction, bringing your development platform incredibly close to full-scale autonomous vehicles.

30.Autoware.universe Multi-Point Navigation Routines

Features pre-configured Autoware. universe multi-point navigation examples. By utilizing Python scripts to dispatch sequential waypoint commands, the system achieves smooth, continuous multi-point autonomous navigation across complex routes.

Frequently Asked Questions

What is the LSR100 designed for?

The LSR100 is designed for indoor service-robot research, ROS development, SLAM and navigation, embodied AI, computer vision, semantic interaction, autonomous mobile robot development, education, and secondary development.

What type of chassis does the LSR100 use?

It uses a differential-drive chassis with brushless hub drive wheels, 2-inch caster wheels, and pendulum suspension.

What are the payload, maximum speed and endurance?

payload capacity below 60 kg, maximum speed up to 2.4 m/s, and approximately 6.5 hours of light-load endurance.

Which onboard computers are available?

Controller options include Raspberry Pi 5 (8 GB), Jetson Orin Nano Super (8 GB), and Jetson Orin NX Super (16 GB), depending on the selected product version.

Which ROS versions are supported?

Raspberry Pi 5 supports ROS 2 Humble, while NVIDIA Orin configurations support ROS 1 Noetic plus ROS 2 Humble.

What LiDAR options are available?

The platform supports 2D TOF LiDAR options including N10P and M10P. The Ultra version includes a multi-line LiDAR, and 16-line or 32-line multi-line LiDAR can be configured for other versions by request.

Does the LSR100 support 2D and 3D SLAM?

Yes. The development resources include 2D LiDAR mapping, RTAB-Map 3D mapping, ORB-SLAM2, RGB-D mapping, and LiDAR–vision fusion workflows.

Does it support Nav2 and autonomous exploration?

Yes. Navigation resources include Nav2 controller examples, RRT autonomous exploration, path recording and playback, multi-point navigation, coverage planning, and obstacle-avoidance workflows.

Can the LSR100 run large language models or multimodal AI?

The development resources support local or cloud-connected LLM integration and multimodal AI workflows. Practical model size, latency, and real-time performance depend on the selected controller and software configuration.

Is a voice module required for AI voice interaction?

Yes. Large-model voice interaction and multimodal voice functions require the corresponding voice module.

Does the robot support semantic navigation?

Yes. Development resources include semantic mapping, semantic navigation, and voice-based semantic labeling workflows when the required AI, voice, mapping, and compute modules are configured.

Can I use the LSR100 with Autoware?

Development resources include Autoware-oriented outdoor mapping and multi-point navigation examples. An NVIDIA Orin-series controller with 16-line or 32-line LiDAR is recommended for this workflow.

Does the LSR100 support autonomous charging?

An optional autonomous charging workflow is available. The corresponding development resources cover charging-station coordinate calibration, docking methods, and low-battery return logic.

Can I add 4G or 5G remote communication?

Yes. Optional 4G / 5G remote modules support long-range video transmission and remote control over cellular networks.

What open-source development resources are included?

Development resources cover ROS, SLAM, navigation, URDF models, perception, YOLO object detection, visual and LiDAR tracking, path planning, autonomous exploration, mobile control tools, wireless code flashing, voice interaction, and secondary development. Availability of individual functions depends on the installed hardware and selected configuration.