MH3 (Mirsee Robotics) vs Dex (Richtech Robotics)

MH3 is taller (170 x 50 x 40 cm) and heavier (75 kg), indicating a larger structural footprint and a build optimized for stability and human-like reach, with force sensors integrated in the arms for precision tasks. Dex is more compact (137 x 50 x 45 cm) and lighter (55 kg), prioritizing a smaller footprint and easier deployment in constrained environments. MH3’s dimensions suggest suitability for heavy-duty manipulation and tool handling, while Dex’s reduced mass supports faster installations and lower floor loading. Both platforms present humanoid form factors but trade off size and payload capacity against deployability.

MH3 specifies a maximum walking speed of 3.5 km/h (0.97 m/s) and supports indoor SLAM, visual SLAM and LiDAR mapping for navigation, indicating multi-modal localization and mapping capability. Dex lists a marginally higher max walking speed of 3.6 km/h (1 m/s) and uses visual SLAM with depth sensing plus indoor SLAM, with LiDAR mapping as optional, showing a perception stack oriented toward depth-assisted visual navigation. MH3’s LiDAR-first mapping as standard implies stronger out-of-the-box performance in structured indoor mapping, whereas Dex’s optional LiDAR keeps weight and cost down while providing depth-assisted autonomy. Both offer teleoperation and autonomous modes, enabling mixed manual/autonomous deployments.

MH3 includes RGB cameras, stereo cameras, LiDAR, ultrasonic sensors, IMU, gyroscope, arm force sensors and temperature sensors, forming a broad sensor fusion suite for precise perception and manipulation. Dex provides front RGB cameras (2), depth cameras (2), two arm-mounted cameras, IMU, gyroscope, force/torque arm sensors, ultrasonic obstacle detection and temperature sensors, emphasizing depth sensing and localized arm views. MH3’s stereo + LiDAR combination supports dense 3D mapping and long-range sensing, while Dex’s multiple depth cameras and arm-mounted views focus on workspace perception and manipulation context. Both platforms include force sensing in the arms for safe contact tasks and manipulation feedback.

MH3 supports autonomous task programming and learned behaviors alongside immersive teleoperation with proprietary teleoperation software, running on a ROS2-based OS with C++ and Python APIs for custom autonomy stacks. Dex offers autonomous operation, teleoperation and learned behaviors via reinforcement learning, plus proprietary AI modules layered on a ROS2-based OS with Python and C++ APIs, indicating a stronger emphasis on on-board learning workflows. MH3’s control suite highlights VR-based operator-in-the-loop control for precision tasks, while Dex emphasizes reinforcement-learned behaviors for adaptive autonomy in varied environments. Both provide developer-accessible APIs for integration with external AI and control systems.

Both MH3 and Dex list a battery usable life of 4 years, indicating similar expected battery lifecycle before replacement or refurbishment. No runtime-per-charge figures are provided for either platform, so direct comparisons of operational hours or energy consumption cannot be made from the provided data. MH3’s heavier mass (75 kg) and LiDAR as standard may increase power draw relative to Dex’s lighter 55 kg platform with optional LiDAR. Battery lifecycle parity suggests manufacturers expect similar maintenance cadences even if run-time profiles differ by payload and sensor usage.

MH3 targets skilled labor augmentation in manufacturing assembly, remote operation in hazardous sites, precision tool handling and service tasks requiring human-like dexterity, aligning with its larger size, force-sensing arms and immersive teleoperation capabilities. Dex targets industrial automation, warehouse logistics, customer service, healthcare assistance and R&D, matching its lighter frame, depth-sensing emphasis and reinforcement-learning support for varied task sets. MH3 fits deployments where remote operator control and heavy-duty precision matter, while Dex fits multi-purpose commercial settings that prioritize mobility, depth perception and lower mass. Both platforms support collaborative modes and are suitable for human-shared workspaces with appropriate safety integration.

Both robots run a ROS2-based operating system and expose C++ and Python APIs, facilitating integration with standard robotics toolchains and custom development. MH3 includes proprietary teleoperation software focused on immersive VR and haptics, providing specialized tooling for operator-in-the-loop workflows and precision remote control. Dex includes proprietary AI modules and explicit reinforcement learning support, indicating built-in ML tooling for behavior learning and adaptation. The shared ROS2 base eases portability of algorithms between platforms, but proprietary modules mean vendor-specific integration work will differ per deployment.

MH3 is priced at $250,000 - $350,000, reflecting a higher-cost, larger platform with integrated LiDAR and immersive teleoperation features. Dex is priced at $150,000 - $250,000, offering a lower entry price and a lighter, more compact form factor with depth-focused sensing and optional LiDAR. The price ranges overlap, allowing procurement decisions to weigh required features—precision teleoperation and full LiDAR on MH3 versus Dex’s lower mass, depth cameras and reinforcement-learning focus—against budget constraints. Total cost of ownership will also depend on optional sensors, software licenses and deployment-specific integration effort.

Both MH3 and Dex list force limiting, collision detection and emergency stop functionality as core safety provisions, and both support collaborative modes with proximity or compliant sensors for shared human workspaces. MH3 emphasizes collaborative mode with safe proximity sensors alongside arm force sensors, suitable for close-proximity skilled tasks. Dex specifies force-limiting joints and collision detection sensors and notes compliance with safety standards for collaborative operation. Shared safety feature sets indicate both platforms are engineered for human-adjacent deployments, with differences likely in sensor placement and certification details not provided here.