/VLMRobotics

Vision Language Models for Robotics

VLM Robotics - Vision Language Models for Robotics

Models

QuadRuped Comparison

Feature Unitree Go2 Anybotics Anymal Tencent Max
Weight 12 kg 50 kg 14 kg
Battery Capacity 8,000 mAh (15,000 mAh optional) 932.4 mAh Not specified
Max Speed Not specified 1.3 m/s 0.5 m/s
Dimensions Not specified 930 x 530 x 890 mm Not specified
Ingress Protection Not specified IP67 Not specified
Special Features 4D LIDAR, AI Mode, Various Actions Autonomous Navigation, AI-based Mobility Lifelike Agility, Multi-agent Chase Tag Game
Key Technologies Ultra-wide 4D LIDAR, Advanced AI Mode, ISS2.0 Intelligent Side-follow System Complete autonomous navigation, High-resolution sensors, Machine learning Pre-trained latent representations, Neural controller with information bottleneck structure
Applications Various poses and actions, Obstacle avoidance Industrial inspections, Complex environment navigation Research, Agility demonstrations, Multi-agent games

Some notable points about each robot:

  1. Unitree Go2: Features advanced AI capabilities and a 4D LIDAR system for improved recognition. It has a high battery capacity and can perform various actions like jumping, stretching, and sitting down[1].

  2. Anybotics Anymal: Designed for industrial use with autonomous navigation and AI-based mobility. It has IP67 protection, making it suitable for harsh environments. The robot can climb stairs, navigate narrow spaces, and operate in slopes up to ±30°[2].

  3. Tencent Max: Focuses on lifelike agility and is used for research purposes. It can participate in multi-agent games like Chase Tag and demonstrates animal-like behaviors. The robot uses pre-trained neural networks to adapt to various environments[3].

Each robot has its unique strengths and applications, ranging from industrial inspections to research and entertainment. The comparison shows the diversity in quadruped robot designs and capabilities across different manufacturers.

Robotic Arm Comparison

Here is a comparison table of various robotic arms that can be controlled programmatically, including details such as degrees of freedom, cost, and camera resolution:

Robotic Arm Model Degrees of Freedom Cost (USD) Camera Resolution Notable Features
Poppy Humanoid Arm 4 ~500 Not specified Open-source, 3D-printed, suitable for education and research[3].
Universal Robots UR5 6 ~35,000 Optional add-on Collaborative robot, easy programming, versatile applications[5].
KUKA LBR iiwa 7 ~60,000 Optional add-on Sensitive to touch, suitable for delicate tasks, advanced safety features[5].
ABB IRB 120 6 ~25,000 Optional add-on Compact, lightweight, high precision, suitable for small parts assembly[5].
Fanuc LR Mate 200iD 6 ~30,000 Optional add-on High-speed, compact, versatile, suitable for various industrial applications[5].
Dobot Magician 4 ~1,200 720p Affordable, educational purposes, multiple end-effectors available[5].

Notes:

  • Degrees of Freedom (DoF): Indicates the number of independent movements the arm can perform.
  • Cost: Approximate cost in USD, which can vary based on configuration and additional features.
  • Camera Resolution: Some models offer optional camera modules with varying resolutions.
  • Notable Features: Highlights unique aspects or applications of each robotic arm.

This table provides a snapshot of different robotic arms, focusing on key attributes that are important for programmatic control and various applications.

References:

[1] https://www.reddit.com/r/ControlTheory/comments/f93m4t/control_system_for_robotic_arm/
[2] https://docs.revrobotics.com/duo-control/programming/hello-robot-autonomous-robot/arm-control-blocks
[3] https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5660981/
[4] https://www.researchgate.net/publication/367340136_Implementation_of_a_Mathematical_Model_and_Comparison_of_Control_Algorithms_for_an_Industrial_Robot_Arm
[5] https://www.universal-robots.com/in/blog/types-of-robotic-arms/
[6] https://shop.unitree.com/products/unitree-go2
[7] https://www.anybotics.com/anymal-technical-specifications.pdf
[8] https://tencent-roboticsx.github.io/lifelike-agility-and-play/
[9] https://www.theblifemovement.com/weilan-alpha-dog-if-rolex-made-cyber-dogs/

Reinforcement Learning Libraries

Blog Posts