Speedbot has developed the leading multi-view high precision measurement system in the world for large-size mechanical parts and products. Multiple patented technologies form the foundation of the system along with the most advanced 3D vision technologies and multi-view metrology computations including high-precision multi-camera joint calibration, multi-view alignment of 3D CAD models, robust feature detection on weak textured surfaces and super-high-resolution multi-view stereo and 3D reconstruction. This allows the Speedbot system to measure objects in a fraction of the time it would take laser scanning and structured light imaging.

The system can measure over a thousand points of inspection on an object with submillimeter precision of 0.2mm in under 100s on a part 2m in size, including planes, holes and other complex points where laser scanning and other technologies fall short or structured light imaging has a limited field of view.

Many of Speedbot’s products are based on a proprietary platform SpeedHub which provides “one-click” hand-eye calibration, multi-camera calibration, heterogeneous sensor joint calibration and vision-inertial calibration.

Speedbot has also developed proprietary techniques based on multi-modal visual perception and deep feature fusion for vision-based localization and orientation. This enables our highly accurate and robust object pose estimation by the robots – even in situations where objects are occluded by other objects, there are complicated textures or there are lighting condition changes.

Speedbot technology also allows for real-time, dense RGBD reconstruction on a fast moving camera (1~3m/s) which enables fast and high-quality mapping, localization and navigation on fast moving robots.

Accurate and fast object detection and recognition is one minor and one of the most important attributes in any automation system. Speedbot features state-of-the-art algorithms for fine-grained detection, segmentation and recognition of products against any number of products. These algorithms are built upon advanced recognition techniques such as multi-modal visual information fusion and hierarchical visual representation and learning which results in exceptionally high object identification using 2D and 3D vision. Several Speedbot algorithms are the best in class in terms of performance.

The speed and accuracy of the Speedbot system is further augmented by proprietary built annotation-rich datasets to detect defects and a series of high-performance and nimble detection algorithms using a deep neural network to train the system. Speedbot also adopts cutting-edge techniques in neural architecture search to allow for the fast development of inspection algorithms for a variety of surface materials, defect types, and lighting conditions.

This is not technology you have seen before.

Ship less air with Speedbot’s optimized packing technology.

Speedbot’s intelligent planning and programming uses deep reinforcement learning techniques to enable the robot to attain human level performance in maximizing packing space and minimizing decision making time. Speedbot’s proprietary motion planning and obstacle avoiding algorithms and software allow robots to collaborate together to pick and pack objects safely without colliding. These algorithms also allow for companies to put restrictions on certain SKU’s such as restricting the placement of more fragile objects to the top of a pallet.

Precision grasping has been a challenge for robotic manipulation of objects. Speedbot proprietary grippers combined with the latest vision guided control algorithms and software enable hand-eye coordination with exceptional accuracy.

For the logistics and manufacturing industries, the Speedbot proprietary vacuum gripper uses the high precision calibration and control algorithms to ensure gentle but accurate gripping of grocery and produce products, picking products up and placing them on the shelf perfectly every time – all while avoiding shelf walls and other obstacles.

For heavy industry and other manufacturers picking heavy steel parts, Speedbot’s proprietary electromagnetic dot array gripper uses the most advanced control software to set the optimal on-off configuration of the dots for picking steel blanks or parts without picking the waste material or neighbouring parts. Recognition is in the submillimeter level of precision even in a dusty environment.

For automated assembly of parts, Speedbot’s vision guided intelligent path planning enables accurate assembly of small to very large parts (0.1m – 1.5m).