Tencent announced the latest research progress of robots “Roller Boy” Ollie has a sense of touch

A few days ago, Tencent Robotics X Robotics Laboratory announced the latest research progress of its wheel-legged robot Ollie, showing the first exposure of “tactile interaction” and exclusive “two-wheel stepping”, which further enriches the robot’s “senses” and improves its motion control ability. It demonstrates Tencent’s leading layout and continuous exploration in the cutting-edge technology of Robots.

Tencent announced the latest research progress of robots “Roller Boy” Ollie has a sense of touch

 Add tactile interaction to challenge the task of carrying the balance ball above the head

Ollie is a wheel-legged robot self-developed by Tencent. It is also another innovative exploration of Tencent Robotics X Robotics Laboratory after the robot dogs Jamoca and Max. , Human-computer interaction and other fields have made key breakthroughs. With his flexible body, Ollie is also known as the “Roller Boy”.

Tencent Robotics X robotics laboratory has innovatively combined the tactile sensor with the wheel-legged robot Ollie. The sensor is jointly developed by Tencent Robotics X Laboratory and Tsinghua University. It uses a new type of piezoresistive material with ultra-high sensitivity, ultra-large range, ultra-fast response speed, and ultra-strong cycle stability, combined with customized electrode adaptation modules and high-speed signals. The acquisition module is added with a self-developed integrated software and hardware solution, allowing the robot to perceive extremely subtle pressure changes on the body surface. Relevant research results have been accepted by ACS NANO, the top international journal in the field of nanotechnology.

After adding a new type of tactile sensor, Ollie can use his “skin” to sense the contact information of the outside world, including the perception and recognition of the touch method, touch force, touch orientation, touch track shape, and respond in different ways . With tactile support, combined with stable movement capabilities, Ollie can also challenge the difficult tasks of head balance and handling spherical objects, making full use of the contact information between the sphere and the robot surface, combined with the data of its own attitude sensor and joint motor encoder , to achieve the perfect combination of upper body object manipulation ability and lower body movement balance ability.

When a mobile robot is used to complete a handling task, it generally chooses a handling object with a stable contact surface, such as a square object such as a cardboard box; if it is necessary to use a flat “head” to carry a spherical object that is easy to roll and slide, it is much more difficult for the robot. Ollie uses a new type of tactile sensor to perceive the relative position and motion state of the ball, and processes the data in real time to control itself to achieve balanced and stable driving on different terrains, and to keep the ball from falling. The leading technology behind it has laid a solid foundation for mobile robots to improve their own movement and object manipulation capabilities in complex scenes.

Start two-wheel walking, walking on different ground “like walking on flat ground”

The latest Ollie also shows a two-wheel stepping action. Compared with the “stepping in place” action shown last year, it has added the ability to step and move at the same time. It walks on different ground “like walking on flat ground”, and it lasts longer and moves more. Smoother and more stable overall performance.

The completion of two-wheeled step action depends on the action generation technology. This requires dividing the action of the robot into a single-wheel support phase and a double-wheel support phase. The two-wheel support stage is an instantaneous switching state in the stepping action. Although the time is very short, it is very difficult. At this stage, there is relative sliding between the two wheels of the robot and the ground. This complicates the contact forces and dynamics of the robot. Therefore, the research team adopted a data-driven approach, using reinforcement learning methods for the generation of joint angle sequences in this action, and using the generated joint angle sequences for robot action and attitude control.

In the future, Ollie will continue to serve as an experimental platform for cutting-edge technology exploration of Tencent Robotics X Robotics Laboratory, undertaking research tasks in many fields from robot body design, system integration to extensive perception and control planning algorithms, and completing the accumulation of technical capabilities.

The Links:   SGAGS-761KA2A-YR61 3HNA015202-001

Pre:    Next: