Tencent's Robot Dog Performs Chinese Kung Fu and Challenges the Internet Celebrity Boston Robot Dog
Today, the four-legged mobile robot dog Jamoca, jointly developed by Tencent Robotics X Lab and Tencent AI Lab, was unveiled. In addition to the basic skills of walking, running and jumping, it can also walk on plum blossom piles, which makes it look like a dog that has learned martial arts since childhood.
You’ve seen robots that can climb stairs, do somersaults, and dance, but have you ever seen a robot that can walk on plum blossom poles?
Just today, the four-legged mobile robot Jamoca developed by Tencent made its first public appearance. Because it looks like a dog, people also call it a robot dog.
This Jamoca is the first robot dog in the country that can complete the complex challenge of walking on the plum blossom piles. It walks on the plum blossom piles accurately and steadily. With a little practice, it may become a great martial arts master in the future.
Overcoming two major difficulties and taking down the plum blossom pile with technology
The plum blossom pile originated in the late Ming Dynasty and is one of the important training props for Chinese Kung Fu. Practicing the plum blossom pile is also the basis for practicing Qinggong in martial arts novels.
Now, the plum blossom pile has welcomed a new "practitioner" - Tencent's robot dog Jamoca.
It is not easy for humans to master the plum blossom pole, and it is even more of a challenge for robot dogs. Walking on the plum blossom pole tests its mobility in two main ways:
- Understanding the arrangement of plum blossom piles (including location and altitude);
- Choose the best foothold and route, and walk steadily and accurately (Land on the center point of the plum blossom pile).
Jamoca was able to overcome these two major challenges. Its secret is mainly two core technical modules:Perception, motion planning and control modules, as well as whole system design and construction modules.
With the support of these two core technologies, Jamoca can remain calm and confidently cross the plum blossom piles of varying spacing and heights when faced with them.
Self-developed robot control technology gives Jamoca autonomous positioning and obstacle avoidance capabilities
As this video shows, Jamoca completed the difficult challenge, climbing the 60-centimeter-high steps with a 20° slope, and successfully crossed the randomly arranged plum blossom piles one step at a time.
Judging from the results of the Jamoca challenge, its performance has reached extremely high technical standards: perception and positioning with an error within 1 cm, 10 millisecond route planning based on the environment, 1KHz real-time torque control based on dynamics, landing error within 1 cm of the center point of the plum blossom pile, and high coordination of the entire system.
It’s also a robot dog, so why is it so excellent?
Compared with other previous international quadruped robots walking on wooden blocks, the plum blossom pile that Jamoca challenged this time has a smaller foothold area, a higher height, and a combination of steps, making it more difficult to achieve.
How did Jamoca develop such excellent skills? It all started with three basic skills.
1. Accurately perceive the environment
According to the introduction,Jamoca innovatively implements robust eye-foot calibration and uses RGB-D cameras to perceive the surrounding environment in real time.
Jamoca’s own motion trajectory is tracked online through feature point matching, and the vision-based positioning information is fused with the kinematics-based odometer information and IMU (Inertial measurement unit) data to improve the accuracy and frequency of positioning and tracking.
At the same time, algorithms are used to identify and extract the surface area and center point positions of the steps and plum blossom piles, and the positioning and identification data are fused to reconstruct the entire three-dimensional motion scene.
2. Optimal route planning
Based on the perception system, Jamoca can perceive the position information of the robot body and the plum blossom pile in real time.
also,Based on center of mass dynamics, online center of mass trajectory generation and landing point planning are realized.
Therefore, while ensuring that its four legs do not miss a step in the next step, it can also optimize a motion trajectory with the shortest moving length and the lowest overall energy consumption, and can perform the above motion planning online and continuously.
I have to say, the robot dog has a lot of little plans in mind.
3. Accurate and real-time force control, walking, running and jumping freely
Based on real-time body state estimation, Jamoca can combine the center of mass dynamics model to construct optimization problems, solve the ground reaction force required by the robot foot in real time, and combine feedback control to achieve accurate and robust real-time force control. Therefore, walking, diagonal trotting, and jumping are all easy for Jamoca.
At present, Jamoca is mainly used for scientific research experiments in the laboratory. In the future, its online environment perception, optimal motion planning and real-time motion control capabilities will help Tencent's other robot products better adapt to complex real-world environments.
News Source:
Tencent AI Lab:https://mp.weixin.qq.com/s/vWAKLhOMRDehppXJG2Pktw
-- over--