KAIST’s DreamWaQ++ robot uses vision, LiDAR and AI to navigate terrain like animals in real time.
A KAIST research team has developed a quadrupedal robot control system that allows machines to walk through complex terrain by understanding their surroundings in real time using cameras, LiDAR, and onboard sensors, enabling more adaptive movement similar to how animals adjust their steps in unfamiliar environments.
The system, called DreamWaQ++, builds on earlier work that enabled robots to walk without visual input by relying only on internal sensors such as joint encoders and inertial measurement units. While that earlier approach allowed stable movement even in low-visibility conditions, it could only react after physical contact with obstacles, limiting its ability to avoid hazards in advance.DreamWaQ++ adds external perception to the system, combining vision and depth sensing with proprioceptive feedback. This allows the robot to identify obstacles before reaching them and adjust its walking strategy in real time. The result is a shift from reactive movement to perception-based locomotion, where the robot actively interprets its environment while moving.The researchers say the upgrade makes the robot more robust in unpredictable environments such as disaster zones, industrial sites, and uneven natural terrain.Perception meets movementTo achieve this, the team developed a multimodal reinforcement learning framework that processes different sensor inputs simultaneously while maintaining lightweight computation for real-time control. The system can also switch between sensing modes when errors occur, improving stability and adaptability across conditions.In tests, the robot showed strong performance across multiple challenging scenarios. It climbed a 50-step staircase covering 30.03 meters horizontally and 7.38 meters vertically in just 35 seconds, outperforming both blind locomotion systems and existing perception-based controllers.In steep terrain experiments, it successfully climbed slopes of up to 35 degrees, significantly steeper than its training conditions, while reducing motor load compared to previous methods. It also demonstrated adaptive decision-making by selecting efficient paths without external planning systems.Learning like living systemsThe robot also showed exploratory behavior in uncertain environments, pausing to assess drop-offs before moving forward. In obstacle trials, it cleared barriers taller than itself while carrying additional payloads, showing strong balance and stability under load.The system’s training allowed it to generalize beyond its original conditions. Although trained on relatively small obstacles, it achieved high success rates on much larger real-world structures, indicating strong adaptability rather than simple pattern repetition.Researchers believe the approach could extend to other robotic platforms, including wheeled-legged and humanoid systems, expanding its use in inspection, agriculture, forestry, and emergency response operations.“This research shows that robots have advanced beyond simply moving to a level where they understand the environment and make decisions on their own,” said Professor Hyun Myung, who led the study. “We will further expand this into intelligent mobility technologies applicable in various real-world environments.”The study has been published in the journal IEEE Transactions on Robotics.
Dreamwaq++ KAIST LIDAR Quadruped Robot Reinforcement Learning Robotics Terrain Adaptation
United States Latest News, United States Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
USA Network's 4-Part Psychological Thriller Series Keeps Getting BetterPortia Doubleday as Angela in Mr. Robot
Read more »
Chatty robot helps seniors fight loneliness through AI companionshipSome American senior citizens have a new companion to help them combat loneliness: a robot. ElliQ is a tabletop device that uses artificial intelligence to conduct humanlike conversations with its owners. Its creators have designed the device for senior citizens who live alone.
Read more »
Only 5 Animated Movies in the 2020s Can Be Considered True MasterpiecesDog and Robot walking together while holding hands during the cold, winter season in Robot Dreams
Read more »
More than 100 robots showcased in Hong KongA humanoid robot called X2 Ultra from China's leading humanoid robot manufacturer AGIBOT has been impressing visitors in Hong Kong. The robot is about the size of a primary school student. It sang and conversed in Mandarin and English on Monday. The robot was part of an exhibition featuring more than 100 robots.
Read more »
Toyota’s new CUE7 robot uses smart vision and control to make precise shotsToyota unveils CUE7, advancing AI basketball robot with improved sensing, planning and precision beyond earlier record setting precedessors.
Read more »
First humanoid robot with embodied intelligence for high-risk tasks enters serviceChina has reportedly deployed its first 'embodied' intelligent humanoid robot designed for high-risk industrial operations.
Read more »