TruPhysics Robotics Artificial Intelligence
TruPhysics design simulation software solutions to support the robotic development which controls programs by using a real-time and high-resolution physics simulation.
Conventional Approach
A traditional way to teach new objects for grasping positions manually for each situation.
TruPhysics Robotics Artificial Intelligence
Neural Networks learn the best grasping approach in simulation for known objects. Once trained, they can even generalize to near optimal grasps for similar objects encountered in the future.

It is hard to compute the grasping positions in uncertain environments and for handling uncertain objects. The complexity increases if the gripper has more than one joint (e.g. parallel gripper), so various grasping strategies are available. The computed results for a grasping point with grasping strategy has in most cases not more than 80% of grasping success. Moreover, the environment does not allow to reach the grasping points, because of obstacles only standard approach of shaking the system or pushing the obstacles away to create another arbitrary situation.

With TruPhysics the grasping strategies can be validated in the physics-based simulation environment and the results can be fed in AI based neural networks. In addition, the object can be trained with reinforcement learning techniques with the optimal interaction. We create a perfect grasping strategy for each situation, independent of any obstacles. TruPhysics increases the success of every handling operation by approximation of best practices from the training data from the past.

Current projects

Durch die weitere Nutzung der Seite stimmst du der Verwendung von Cookies zu. Weitere Informationen

Die Cookie-Einstellungen auf dieser Website sind auf "Cookies zulassen" eingestellt, um das beste Surferlebnis zu ermöglichen. Wenn du diese Website ohne Änderung der Cookie-Einstellungen verwendest oder auf "Akzeptieren" klickst, erklärst du sich damit einverstanden.