China’s A.I. Powered Robot Nearly Fights Back After Being Hit, Pulled By Trainer 

Please note that this is not intended to be investment advice. The author does not hold any positions in the mentioned stocks.

LimX Dynamics, a Chinese robotics technology firm, recently demonstrated the remarkable capabilities of their Biped Robot P1. During a test, the robot effectively responded to threats from a human and navigated through challenging terrain. According to the company, the robot utilizes reinforcement learning to react to external stimuli, including moving objects and obstacles on its path. This technique is part of a larger set of methods used to train artificial intelligence algorithms. The latest advancements from Zhuji Dynamics, as the firm is also known, further solidify their continuous progress in the field of robotic technology.

LimX Dynamics’ Biped P1 Robot Masters Mountainous Forest Terrain Using Reinforcement Learning

LimX, a robotics company, has a product portfolio that includes a humanoid robot, a quadruped robot, and the biped P1. Recently, the company conducted a test run for the P1 platform in the Tanglang Mountain in Shanghai, China. The aim of this test was to evaluate the platform’s performance in an unpredictable environment with rough terrain. The P1 was required to constantly be aware of its surroundings and use reinforcement learning to incorporate data into its responses.

In a Medium post, LimX revealed information about the test. The company stated that the test’s most challenging aspect was the lack of uniform steps or ground level in the mountainous terrain, unlike urban environments. This test marked the third for the company’s full product portfolio, following previous tests with humanoid and quadruped robots. The quadruped X1 underwent a slow navigation around a steel factory in South China in October of last year.

For its latest test, the company once again released a video showcasing the P1’s journey through various terrains. The test demonstrated the robot’s ability to ascend grassy hills, navigate around obstacles, recover from stumbles, and withstand being kicked, pulled, and hit with a stick without losing its balance. Upon closer examination of the footage, it was observed that the stick used by the trainer actually broke upon impact with the robot’s legs.

Despite not disclosing the distance covered during its first mountainous run, LimX Dynamics did reveal some information about the robot’s training using reinforcement learning.

Reinforcement learning sets itself apart from other branches of machine learning, such as supervised and unsupervised learning, which use or disregard data labels to train software. Instead, it utilizes rewards or penalties to allow a system to make decisions. Although machine learning systems differ from artificial intelligence systems in their ability to learn autonomously, they are still considered a part of the larger realm of A.I.

LimX stated that despite not utilizing any data from its environment, the test was still successful in guiding the P1 through its intricate surroundings. The company has plans to implement similar techniques in their humanoid robot, the CL-1, which recently completed a test in December. During this test, the CL-1 skillfully ascended stairs and maneuvered through various indoor and outdoor obstacles, utilizing data from its surroundings to do so.

The video below features the mountain test of P1 for you to watch.

The provided link is for a video with the URL https://www.youtube.com/watch?v=UpNid_rWDnI.