Motion Stream Interface for Physical AI
The Motion Stream Interface, or MSI for short, enables selected Weiss Robotics servo grippers to be controlled in real time. External controllers, AI systems, or teleoperation solutions can transmit position and force setpoints as a continuous motion stream while directly evaluating feedback data from the gripper. This transforms the gripper into an active component of modern Physical AI architectures — enabling adaptive gripping strategies, advanced bin picking, visual servoing, and teleoperation.
MSI is the interface for reactive systems that do not simply execute gripping motions, but make decisions, apply corrections, and learn while in motion.
APPLICATION AREAS
MSI is designed for applications in which the gripper does not simply execute a command, but dynamically responds to external data. AI models, vision systems, or teleoperation solutions can influence finger movement in real time and integrate the gripper state directly into their control strategy.
Physical AI & Advanced Bin Picking
For AI-driven gripping strategies in which the gripper reacts to sensor data, object position, and model decisions while in motion. MSI enables adaptive finger movements for unstructured, complex, and highly unpredictable part scenarios.
Teleoperation & Remote Manipulation
MSI enables gripper movements to be transmitted as a motion stream and linked with feedback data from the gripper. This allows the implementation of master-slave setups, remote manipulation, and intuitive control concepts with precise real-time tracking of finger movements.
External Gripper Control & Research
For teams developing their own control strategies, prototypes, or experimental gripping algorithms. MSI provides direct access to the gripper’s motion control — for applications such as visual servoing, force-adaptive control, impedance concepts, or learning-based control approaches.
Designed for AI-Driven Robotics
Physical AI is changing the requirements for gripping technology. A gripper no longer only needs to execute predefined movements — it must also be able to respond to object position, process conditions, sensor data, and model decisions.
MSI provides direct access to motion control for this purpose. Motion and force profiles are generated in the external system — for example within an AI pipeline, a vision controller, or a teleoperation environment — and transmitted to the gripper as a continuous stream. Internally, the gripper interpolates the received base points to its 1-ms control cycle while simultaneously providing feedback on position, velocity, force, status, and buffer state.
The Gripper as a Real-Time End Effector for Physical AI
In Physical AI applications, the gripper is more than just a tool mounted on the robot flange. It is the active point of contact between model decisions and the physical world.
With MSI, AI-based systems can generate, evaluate, and adapt gripping motions online. Instead of moving toward a fixed target position, the external controller can continuously modify finger movement — depending on camera feedback, force progression, object behavior, or policy output. This enables learning-based gripping strategies in which the gripper is directly integrated into the decision-making and control loop.
Real-Time Correction for Demanding Bin Picking Applications
In advanced bin picking, conditions often change during the gripping process: objects become jammed, slip, remain partially occluded, or behave differently than predicted by the vision system. Conventional gripping commands quickly reach their limits in such scenarios.
With MSI, an external system can continuously adjust position and force setpoints during execution — for example based on 3D vision, tactile data, grip force estimation, or process feedback.
This makes MSI particularly valuable for bin-picking applications where not only the target pose matters, but also the adaptive motion required to achieve a stable grip.
Transmitting Gripper Motion in Real Time
Teleoperation places high demands on responsiveness, synchronization, and feedback. MSI can transmit gripper movements as a motion stream while simultaneously returning status and feedback data from the gripper.
This enables master-slave scenarios, remote manipulation, and intuitive control concepts in which one gripper defines the motion and another gripper follows it almost in real time. The reference manual describes an example teleoperation setup with two grippers: a master gripper provides position feedback, while a slave gripper tracks and reproduces the movement.
Streaming Instead of Single Commands
MSI follows a stream-oriented control model. The gripper does not generate the trajectory itself. Instead, the external controller continuously sends base points containing target positions and force limits. These base points are internally interpolated by the gripper to its 1-ms control cycle.
As a result, the motion logic remains entirely within the external system. Policies, vision algorithms, control models, or teleoperation systems can directly determine gripper movement without being constrained by a rigid command-based architecture.
Feedback for Adaptive Control Loops
MSI continuously provides feedback data from the gripper. This includes finger position, velocity, motor current, estimated gripping force, status flags, timestamps, operating state, and buffer fill level.
External systems can use this data to monitor movements, classify gripping states, detect contact phases, or adapt control strategies online. MSI therefore serves not only as a motion execution interface, but also as a real-time monitoring interface for data-driven robotic applications.
For Teams That Want Full Control Over Gripping
MSI is designed for development teams that want to integrate the gripper deeply into their own control architecture — typically in areas such as Physical AI, advanced bin picking, teleoperation, visual servoing, and adaptive force control.
AI models can generate motion and force profiles directly within the process. Vision or process data can correct finger movements during execution. Teleoperation systems can transmit gripper movements and feedback da
Developing Physical AI or Teleoperation Systems?
With the Motion Stream Interface, you can integrate the gripper directly into your real-time architecture — from AI-based bin picking and adaptive gripping strategies to master-slave teleoperation.
Talk to us about your application. We support you in selecting compatible servo grippers and evaluating the technical requirements of your MSI integration.
Tel.: +49 7141 94 702-10
Email: sales@weiss-robotics.com
Available Products
WPG series
0,00 €/pcs.
Documentation
Python library and example programs
Wireshark dissector for the MSI protocol
Maximum Control — with Responsibility in the External System
MSI is designed for demanding real-time applications. The external controller is responsible for trajectory planning, timing, buffer management, and error handling. The gripper provides internal execution, interpolation, and feedback. This architecture offers maximum freedom for custom control algorithms, but it also requires careful validation of motion profiles. Especially in highly dynamic movements, rapid direction changes, or continuously high force applications, factors such as mechanical load, temperature, duty cycle, and safety functions must be taken into account.