


Super generalization ability makes large models a ray of hope for "general artificial intelligence".
However, reading thousands of books is not as good as traveling thousands of miles. In an open environment, large models need to truly "walk" into the physical world in order to truly understand complex tasks and solve practical problems.
Recently, Professor Li Xuelong’s team conducted innovative research on autonomous drone swarms in an open environment. They used domestic large-scale models to successfully realize human-computer and multi-machine dialogue interaction in an open environment, breaking the interaction barriers between humans and machines. This research further expands the application scenarios of local security, allowing large drones to soar in real life
Inspired by human cognitive models, our team summarized the highly autonomous cognitive process as " The three-dimensional interaction of "Thinking Computing-Entity Control-Environment Perception" has been established, and a "group chat-style" control framework for autonomous drones driven by the "Scholar·Puyu" open source large model has been established. We equip each drone with an intelligent brain, allowing the drone group to dynamically collaborate through language communication to achieve intelligent interaction, active perception and autonomous control in open environments and complex tasks. This move improves the autonomy of drone mission execution
In general, the main capabilities of autonomous drone swarms include human-like conversational interaction, active environment perception and autonomous entity control
humanoiddialogueinteraction
Figure 1 Drone group chat communication
Explore the interaction between human users and drones, allowing drones to understand complex The user needs in the mission are the prerequisites for realizing autonomous drones.
In response to this, the team proposed a "group chat-style" dialogue interaction method, which converts various information such as sounds, images, and the drone's own status into a natural language dialogue form through a large model, realizing the dialogue between users and Drones, and autonomous and intuitive ways to interact with drones.
In order to improve the stability and safety of complex tasks, the team designed an efficient real-time feedback mechanism. This mechanism enables the drone to report its status through dialogue and seek user confirmation at key nodes in mission execution. At the same time, this mechanism can also greatly improve the efficiency of task execution
Active environment perception
Figure 2 Actively discover and approach the target
Figure 3 Dynamic environment obstacle avoidance
During flight, the drone actively senses the external environment and adjusts the mission plan in real time, which is a key link in completing complex tasks.
In order to deal with this problem, the team developed a task-guided active perception mechanism and proposed multi-sensor fusion low-altitude search, dynamic obstacle avoidance and visual positioning algorithms
In actual mission execution During the process, we can dynamically adjust the flight path and observation posture of the drone based on the perceived information and mission goals. We can try to perceive the world around us from different angles and positions, and gradually reduce the uncertainty in the environment to achieve efficient information collection and task execution
Autonomous Control
Figure 4 Autonomous target grabbing
Figure 5 Heterogeneous drone cluster collaborative control
The key research is to explore the form of composite agents , to enhance its ability to handle complex tasks. In the era of large models, this is a key area for new intelligent agents
In order to solve this problem, the R&D team used the drone platform to design end effectors such as grippers, upgrading traditional drones to " "Flying Robot", endows it with grabbing capabilities
At the same time, a heterogeneous UAV cluster collaborative control mechanism is also established, and combined with environmental perception feedback, the flight status of the UAV formation is adjusted in real time so that the cluster can Division of labor and cooperation to perform tasks such as regional search, target positioning and crawling
The team successfully tried to apply the three-dimensional interaction model of biological intelligence "thinking computing-entity control-environment perception" to autonomous agents, forming a large-scale autonomous drone cluster. This kind of cluster uses large-scale language models, drone platforms and a variety of sensors to achieve conversational interaction, active perception and autonomous control. This technology is of great significance for the application in on-site security scenarios such as security inspections, disaster rescue, and air logistics
References: Li Xuelong, Vicinagearth security, Communications of the Computer Society of China, 18(11) ), 44-52, 2022
The above is the detailed content of NPU launches an innovative UAV control framework: enabling group chat-style interaction, active perception of the environment, and autonomous control of UAVs. For more information, please follow other related articles on the PHP Chinese website!

1 前言在发布DALL·E的15个月后,OpenAI在今年春天带了续作DALL·E 2,以其更加惊艳的效果和丰富的可玩性迅速占领了各大AI社区的头条。近年来,随着生成对抗网络(GAN)、变分自编码器(VAE)、扩散模型(Diffusion models)的出现,深度学习已向世人展现其强大的图像生成能力;加上GPT-3、BERT等NLP模型的成功,人类正逐步打破文本和图像的信息界限。在DALL·E 2中,只需输入简单的文本(prompt),它就可以生成多张1024*1024的高清图像。这些图像甚至

Wav2vec 2.0 [1],HuBERT [2] 和 WavLM [3] 等语音预训练模型,通过在多达上万小时的无标注语音数据(如 Libri-light )上的自监督学习,显著提升了自动语音识别(Automatic Speech Recognition, ASR),语音合成(Text-to-speech, TTS)和语音转换(Voice Conversation,VC)等语音下游任务的性能。然而这些模型都没有公开的中文版本,不便于应用在中文语音研究场景。 WenetSpeech [4] 是

“Making large models smaller”这是很多语言模型研究人员的学术追求,针对大模型昂贵的环境和训练成本,陈丹琦在智源大会青源学术年会上做了题为“Making large models smaller”的特邀报告。报告中重点提及了基于记忆增强的TRIME算法和基于粗细粒度联合剪枝和逐层蒸馏的CofiPruning算法。前者能够在不改变模型结构的基础上兼顾语言模型困惑度和检索速度方面的优势;而后者可以在保证下游任务准确度的同时实现更快的处理速度,具有更小的模型结构。陈丹琦 普

由于复杂的注意力机制和模型设计,大多数现有的视觉 Transformer(ViT)在现实的工业部署场景中不能像卷积神经网络(CNN)那样高效地执行。这就带来了一个问题:视觉神经网络能否像 CNN 一样快速推断并像 ViT 一样强大?近期一些工作试图设计 CNN-Transformer 混合架构来解决这个问题,但这些工作的整体性能远不能令人满意。基于此,来自字节跳动的研究者提出了一种能在现实工业场景中有效部署的下一代视觉 Transformer——Next-ViT。从延迟 / 准确性权衡的角度看,

3月27号,Stability AI的创始人兼首席执行官Emad Mostaque在一条推文中宣布,Stable Diffusion XL 现已可用于公开测试。以下是一些事项:“XL”不是这个新的AI模型的官方名称。一旦发布稳定性AI公司的官方公告,名称将会更改。与先前版本相比,图像质量有所提高与先前版本相比,图像生成速度大大加快。示例图像让我们看看新旧AI模型在结果上的差异。Prompt: Luxury sports car with aerodynamic curves, shot in a

译者 | 李睿审校 | 孙淑娟近年来, Transformer 机器学习模型已经成为深度学习和深度神经网络技术进步的主要亮点之一。它主要用于自然语言处理中的高级应用。谷歌正在使用它来增强其搜索引擎结果。OpenAI 使用 Transformer 创建了著名的 GPT-2和 GPT-3模型。自从2017年首次亮相以来,Transformer 架构不断发展并扩展到多种不同的变体,从语言任务扩展到其他领域。它们已被用于时间序列预测。它们是 DeepMind 的蛋白质结构预测模型 AlphaFold

人工智能就是一个「拼财力」的行业,如果没有高性能计算设备,别说开发基础模型,就连微调模型都做不到。但如果只靠拼硬件,单靠当前计算性能的发展速度,迟早有一天无法满足日益膨胀的需求,所以还需要配套的软件来协调统筹计算能力,这时候就需要用到「智能计算」技术。最近,来自之江实验室、中国工程院、国防科技大学、浙江大学等多达十二个国内外研究机构共同发表了一篇论文,首次对智能计算领域进行了全面的调研,涵盖了理论基础、智能与计算的技术融合、重要应用、挑战和未来前景。论文链接:https://spj.scien

说起2010年南非世界杯的最大网红,一定非「章鱼保罗」莫属!这只位于德国海洋生物中心的神奇章鱼,不仅成功预测了德国队全部七场比赛的结果,还顺利地选出了最终的总冠军西班牙队。不幸的是,保罗已经永远地离开了我们,但它的「遗产」却在人们预测足球比赛结果的尝试中持续存在。在艾伦图灵研究所(The Alan Turing Institute),随着2022年卡塔尔世界杯的持续进行,三位研究员Nick Barlow、Jack Roberts和Ryan Chan决定用一种AI算法预测今年的冠军归属。预测模型图


Hot AI Tools

Undresser.AI Undress
AI-powered app for creating realistic nude photos

AI Clothes Remover
Online AI tool for removing clothes from photos.

Undress AI Tool
Undress images for free

Clothoff.io
AI clothes remover

AI Hentai Generator
Generate AI Hentai for free.

Hot Article

Hot Tools

MinGW - Minimalist GNU for Windows
This project is in the process of being migrated to osdn.net/projects/mingw, you can continue to follow us there. MinGW: A native Windows port of the GNU Compiler Collection (GCC), freely distributable import libraries and header files for building native Windows applications; includes extensions to the MSVC runtime to support C99 functionality. All MinGW software can run on 64-bit Windows platforms.

mPDF
mPDF is a PHP library that can generate PDF files from UTF-8 encoded HTML. The original author, Ian Back, wrote mPDF to output PDF files "on the fly" from his website and handle different languages. It is slower than original scripts like HTML2FPDF and produces larger files when using Unicode fonts, but supports CSS styles etc. and has a lot of enhancements. Supports almost all languages, including RTL (Arabic and Hebrew) and CJK (Chinese, Japanese and Korean). Supports nested block-level elements (such as P, DIV),

WebStorm Mac version
Useful JavaScript development tools

Atom editor mac version download
The most popular open source editor

ZendStudio 13.5.1 Mac
Powerful PHP integrated development environment
