PhD researchers - National University of Singapore
CASE STUDIES
Mobile Manipulation Robot for Advanced Robotics Research
Our Answer
Client
PhD researchers - National University of Singapore
Client Requirements
To push the boundaries of integrated robotics, the lab required a platform capable of:
-
Integrated Perception & Manipulation: A seamless bridge between 3D vision, autonomous navigation, and robotic arm control.
-
Complex Indoor Navigation: Reliable SLAM (Mapping & Localization) and dynamic obstacle avoidance in crowded lab environments.
-
AI & Computer Vision: Object recognition and pick-and-place tasks using high-precision depth sensing.
-
Developer-Friendly Ecosystem: Full compatibility with ROS2 Humble and an Open-Source architecture for secondary development of proprietary algorithms.
-
Rapid Deployment: A "Ready-to-Research" hardware stack to eliminate months of manual integration.
1. Robust Hardware Foundation
-
Advanced Chassis: Features patented chassis technology with an independent suspension system and high-torque hub motors for smooth movement.
-
Collaborative Power: Integrated with the Realman ECO65 Arm, offering 6 degrees of freedom (6DoF) for precise manipulation tasks.
-
Autonomous Endurance: Supports automatic recharging, ensuring the robot is always ready for long-duration experiments.
2. Full-Stack Perception Suite (Multi-Sensor Fusion)
-
Dual LiDAR System: Equipped with two Leishen TOF LiDARs for 360° coverage.
-
3D Vision: Features the Orbbec Gemini Pro binocular depth camera for high-fidelity environment reconstruction.
-
Safety First: A comprehensive ultrasonic anti-collision system provides hardware-level safety during autonomous missions.
3. Open-Source Software & ROS2 Integration
-
Native ROS2 Humble Support: Pre-configured drivers and packages for the latest robotics middleware.
-
Fully Open-Source: Access to all chassis source code, enabling researchers to modify control logic at the lowest levels.
-
Pre-Integrated Algorithms: Out-of-the-box support for SLAM, autonomous path planning, and vision-guided grasping.
Scalable UGV Swarm Formation Platform for Multi-Agent Research
Client
A high-tech Unmanned Systems Company specializing in UGV swarm intelligence and multi-agent coordination
Client Requirements
-
A ready-to-use multi-robot platform to quickly validate formation and coordination algorithms
-
Minimize hardware development time and engineering overhead
-
Support custom control laws and algorithm integration
-
Provide a scalable multi-robot setup for different test scenarios
-
Ensure compatibility with existing development environments (e.g., ROS)
Our Answer
1. Flexible Fleet & Chassis Configuration
-
Scalable Deployment: Supports modular fleet sizes including 3, 5, 10, or 15 units, allowing the client to match specific test scenarios.
-
Diverse Kinematics: Compatible with multiple chassis types, including Mecanum wheels (omnidirectional) and Ackermann steering, covering a wide range of industrial and research applications.
2. Patented "Navigator" Formation Algorithm
-
Advanced Coordination: Built-in "Navigator" formation algorithm enables complex maneuvers like linear queues, triangular arrays, cross-intersections, and X-shaped patterns.
-
Full Source Access: Unlike closed-loop systems, we provided 100% open-source code, allowing the client to modify formation logic, SLAM parameters, and path-planning scripts.
3. Integrated ROS1/ROS2 Development Stack
-
Out-of-the-Box (OOTB): Pre-configured hardware and software environment reduced the client’s setup time from months to days.
-
Seamless Integration: Fully compatible with standard ROS workflows, facilitating easy simulation-to-reality (Sim-to-Real) transitions.
Client
PhD researchers - National University of Singapore
Client Requirements
The client required a heterogeneous unmanned system where a Unmanned Ground Vehicle (UGV) serves as a mobile base for an Unmanned Aerial Vehicle (UAV). Key requirements included providing charging support to significantly extend operational range, high adaptability in complex terrains, autonomous navigation with FPV remote control, and achieving high-precision autonomous landing on a dynamic platform with bidirectional energy and signal transmission.
Our Answer
We proposed an innovative air-ground coupling architecture that defines the UGV as the energy replenishment and mission extension center for the UAV.
-
Core Technology: Integrated an infrared-aligned landing system for dynamic recovery and implemented a nonlinear control strategy based on Lyapunov theory. By decoupling attitude and translational dynamics, we ensured self-stabilizing flight in complex environments.
-
Performance Optimization: Through Time-Varying Model Predictive Control (MPC) and fuzzy logic, task switching efficiency (e.g., hover-to-land) was improved by 40%, and energy consumption was reduced by 22%.
-
Reliable Communication: Employed a hybrid protocol with adaptive topology management, ensuring end-to-end latency below 0.25s and a packet loss rate within 5%.
Client
PhD researchers - ETH
Client Requirements
To address the limitation of attitude recovery in traditional twin-rotor aircraft—caused by fixed centers of gravity (COG) under unstructured disturbances—the client sought an innovative hardware structure and control algorithm. The goal was to achieve efficient and stable transitions across multiple modes, including hover, morphing/tilting, and cruise.
Our Answer
Starting from improved system dynamics, we developed two integrated hardware and algorithmic solutions:
-
Bio-inspired Adaptive Tail: Inspired by the flight mechanics of Odonata (dragonflies), we developed a 3-DOF bio-inspired tail system. By dynamically adjusting the COG, it generates a maximum compensation torque of 0.3N·m, increasing attitude response speed by 35%.
-
Auxiliary Thrust Vectoring: Introduced auxiliary axial vector propellers supporting a 0°–90° thrust vectoring range. Combined with Nonlinear Model Predictive Control (NMPC), this effectively suppressed disturbances in the 0.5–5 Hz frequency band.
-
Intelligent Control Architecture: Utilized Reinforcement Learning (RL) to pre-train flap deflection strategies for various scenarios. This ensures that the aircraft maintains quadrotor-level control robustness during the wing-tilting morphing phase.