Recently, an exciting news came from the field of humanoid robot. In the Beijing Frontier International Artificial Intelligence Research Institute hosted the "walk into the Songyan power" activities, Songyan power founder and chairman Jiang Zheyuan revealed that the company's humanoid robot "small urchin" in Yizhuang humanoid robot half marathon competition won the silver medal, the company's humanoid robot "little urchin" won the silver medal. Market demand surged, only in a month to get more than 1,000 orders, the current total order size has exceeded 2,000 units, the total contract amount of more than 100 million yuan.
Cheolwon Kang gave an in-depth analysis of the application of deep reinforcement learning in the field of robotics and shared the breakthroughs in humanoid robotics technology of Songyan Dynamics at the event. He pointed out that the application idea of Deep Reinforcement Learning mainly consists of four phases: algorithm design, simulation training, sim2sim (Simulation-to-Simulation), and deployment of real machine. This technological framework provides strong support for improving robot motion performance.
According to Cheolwon Kang, Songyan Dynamics has formed two major product lines, including bipedal humanoid robots (N2 and N1) and simulation humanoid robots. These robots not only have excellent motion performance, but also have a wide range of potential applications, aiming to make robots truly accessible to millions of households.
In terms of deep reinforcement learning, Cheolwon Kang explained in detail the math behind it - Markov Decision Process (MDP).He emphasized that through the combination of deep learning and reinforcement learning, the perception and decision-making ability of robots can be significantly improved, thus realizing more efficient and intelligent motion control.
Cheolwon Kang also shared Songyan Power's specific thoughts on deep reinforcement learning applications. He said that the company focuses on applying deep reinforcement learning technology in motion control, grasping, navigation and obstacle avoidance scenarios, and continuously optimizes the performance of the robot through steps such as algorithm design, simulation training, sim2sim validation, and deployment of real machines.
On the scene of the event, Wang Sheng, chairman of Beijing Frontier International Artificial Intelligence Research Institute and partner of InnoAngel Fund, also expressed his views. He believes that the current AI industry is in a stage of rapid development, although the super application has not yet appeared, but with the continuous progress of technology and the increasing maturity of the market, the future is expected to give birth to more disruptive AI applications.
Sheng Wang also pointed out that the body intelligence industry is still in the early development stage, and the technical route has not yet converged. However, this does not mean a lack of market demand. In fact, with the continuous progress of technology and cost reduction, embodied intelligence has already shown a wide range of application prospects in colleges and universities, research institutes and other fields. Some innovative consumer scenarios, such as wedding leasing, have also brought new development opportunities for the embodied intelligence industry.
Songyan Power's new company is located in Beijing's Changping District, which, as one of the important layout areas for Beijing's robotics industry, has already gathered many new types of enterprises and formed a favorable industrial ecology. Jiang Cheolwon said that in the future, Songyan Power will continue to increase its R&D investment in humanoid robotics technology and continuously improve the performance and quality of its products to meet the growing market demand.