Website Home (ChinaZ.com) June 14 News: Researchers from the University of Tokyo have developed a musculoskeletal humanoid robot named Musashi, which is capable of driving small electric vehicles. This robot integrates advanced technology with visual perception and driving capabilities, showcasing its potential in the field of autonomous driving.
Official Demonstration Video
Key Features:
Visual Perception: Equipped with two cameras to simulate human eyes, capable of seeing the road ahead and side mirrors.
Driving Operations: Uses mechanical hands to start and control the vehicle, including turning the key, pulling the handbrake, and turning on the indicators. Operates the accelerator and brake with anti-slip "feet". Precise steering control is achieved through flexible manipulation of the steering wheel with dual arms. Smooth acceleration and braking are achieved using a six-axis force sensor in the feet.
Compliance with Traffic Rules: A learning-based recognition module can identify traffic lights, pedestrians, and other vehicles, making appropriate driving decisions.
Current Limitations of Musashi:
Slow Turning Speed: Only lightly releasing the brake instead of stepping on the accelerator during turns, resulting in turns taking about two minutes.
Acceleration Control Issues: Difficulty maintaining a stable speed during acceleration, especially on roads with varying gradients.
How Musashi Works:
Hardware Design: Aimed at mimicking the human musculoskeletal structure, featuring high flexibility and perceptive abilities, using synthetic fibers as muscles and a sensor system to perceive the environment.
Software System: Combines multiple learning-based modules to achieve autonomous driving and other tasks, including cross-perception network modules, dynamic task control network modules, reflex modules, and recognition modules.
Workflow:
Environment Perception: Obtains environmental information through cameras and other sensors.
Decision and Planning: Generates motion commands based on environmental information.
Execution Control: Performs specific motion operations through the reflex module.
Feedback and Adjustment: Monitors sensor data in real-time to optimize control accuracy.
Application Examples:
The robot can operate the steering wheel and pedals, adjusting driving operations based on visual and auditory recognition results.
Experimental Conclusion:
Musashi demonstrates potential for autonomous driving but needs to improve operational speed, optimize speed tracking performance, and enhance the recognition system's performance under nighttime and complex lighting conditions.
Future work will focus on further integration and optimization of software and hardware systems, outdoor experiments, and the development of new recognition and control technologies.
The results of this research have been published in a paper, and relevant information sources have also been provided with links. The development of the Musashi robot marks a significant milestone in the advancement of autonomous driving technology, foreshadowing a future of more intelligent and automated driving experiences.
Paper Link: https://arxiv.org/pdf/2406.05573