Intelligent robots and intelligent systems （ Professor Zheng Zheng of Dalian University of Technology ）——2. Mobile Robot Perception

1. Mobile Robot Perception

• Working environment and job scene Hierarchical perception
  • Wide area awareness ： On-line 、 large-scale 、 Quickly conduct large-scale modeling and overall research and judgment
  • Local sensing ： real time 、 Refine 、 Support proximity operation
• Research focus
  • Global ： Panoramic perception can also be achieved in a static state
  • all-weather ： Adapt to different weather （ example ： Radar penetration is affected ）
  • GNSS（ GNSS ） Precise positioning under constraints
    • Lightweight （ There is no need to arrange auxiliary equipment such as base stations in advance ）
    • Data density （ Meet the requirements of high-precision map construction and semantic information extraction ）（ Practical accuracy ： Practical application , Because there is generally no large-scale non GNSS, So it's OK to ensure the accuracy of kilometer level ）
  • Multi modal, multi spatiotemporal and heterogeneous perceptual information fusion and matching
    • Data level fusion
    • Feature level fusion
    • Semantic level fusion
    • Decision level fusion
    • Multisensor
  • The adaptability of diversified scene perception based on transfer learning is improved
    • Autonomous perception model is related Real scene simulation software ：Carala, Deepdrive（60%-70% The data comes from highly simulated software , The remaining data is migrated to the real scene by migration learning ）

2. Learning exchange group

• Exchange of learning QQ Group ：725401489