MultiPlantsAI: An AI-driven Personalized Design, Production and Sales System for Plant Landscapes 

MultiPlantsAI is an AI-driven system for the full workflow of plant landscape design, production, and sales. With only text prompts or image inputs, users can quickly generate personalized landscape solutions, visualize them in real spaces, refine the results, and move toward product realization. Centered on a “Text-to-World” experience, the project transforms traditional landscape design from a complex, high-threshold process into an accessible, efficient, and scalable intelligent service. 

Emotion is not a decoration—it is the foundational protocol that underlies human decision-making and collaboration. To accelerate the industrial adoption of AI in healthcare, education, and human-computer interaction, we have engineered a comprehensive multimodal affective intelligence system. Our solution integrates a specialized hardware suite capable of simultaneous non-contact physiological monitoring, facial Action Unit detection, and audio acquisition. By leveraging a cloud-based Multimodal Large Language Model (MLLM) for sophisticated data analysis and dynamic interaction, we are evolving AI from static brains into adaptive minds. This establishes “understanding humans” as a standard capability for the next generation of intelligent systems, paving the way for more empathetic and effective technology. 

The Emotional Intelligence System was awarded the Gold Medal at the International Exhibition of Inventions Geneva in March 2026 

We are developing a 3D-printed biomimetic skin model that integrates physiological layers, vascular networks, and iPSC-derived sensory neurons to restore both structure and perception. By combining bioprinting, stem-cell engineering, neural differentiation, and biomaterials design, this interdisciplinary project addresses the limitations of current grafts—scarring, rejection, and loss of function—and provides a functional platform for regenerative therapy and in-vitro testing. 

Our project addresses the high clinical overhead of traditional electronic facebows by introducing a multi-modal intelligent sensing framework. By integrating vision-radar fusion with wearable technology, we shift the focus from absolute spatial precision to dynamic functional biomarkers captured under daily-life environments. This interdisciplinary approach provides a cost-effective, screening-oriented solution for routine dental practice, enabling the early detection of functional erosion and supporting data-driven clinical decision-making. 

The ambient environment holds enormous potential as both a resource reservoir and an energy bank, offering an important pathway to alleviate global energy shortages and reduce carbon emissions. Humidity and low-grade thermal energy in the environment can both be harnessed for renewable energy production. This project aims to address the dual challenges of excessive carbon emissions and global energy shortages. Centered on a core system for atmospheric water harvesting and photocatalytic hydrogen production, and enabled by three innovative technological approaches, the project covers the entire process from hydrogen generation and detection to separation and purification, efficiently converting these resources into renewable energy such as hydrogen. 

Third Prize at the Provincial Level, The 4th University Electrical and Electronic Engineering Innovation Competition (EEEIC) 

This project develops a dedicated multi-UAV system for post-disaster indoor and outdoor environments. It supports exploration, reconstruction, object search, and scene understanding, while improving the efficiency of disaster response.  

Third Prize for Oral Presentation, The 5th Guangdong Postgraduate Academic Forum - Sub-forum on Intelligent Manufacturing and Robotics. 

Che Z, Zhang Y, Liang S, et al. GUIDE: A Diffusion-Based Autonomous Robot Exploration Framework Using Global Graph Inference[J]. arXiv preprint arXiv:2509.19916, 2025. 

Zhang M, Du Y, Wu C, et al. Apexnav: An adaptive exploration strategy for zero-shot object navigation with target-centric semantic fusion[J]. IEEE Robotics and Automation Letters, 2025. 

Luo Y, Wang Y, Chen H, et al. Fermi: Flexible radio mapping with a hybrid propagation model and scalable autonomous data collection[J]. arXiv preprint arXiv:2504.14862, 2025. 

This project develops a collaborative task execution solution for heterogeneous robots based on quadruped robots, robotic arms, swarm robots and object navigation. It effectively solves the problem of limited grasping range of robotic arms due to their own structure, and lays a foundation for the subsequent exploration of heterogeneous robot collaboration. 

Lumotech is an AI interior design platform tailored for the B-end home furnishing industry. With just interior space images, it delivers magical home design experiences, solving the pain points of unvisualized overall matching effects, frequent revisions and low efficiency in traditional home decoration.  

Leveraging LLM dialogue, 3D reconstruction and AIGC design technologies, it generates interior preview renderings in one minute, enabling holistic space preview, reducing trial-and-error costs, and boosting shopping guide credibility as well as users' purchase intention. 

MicroBonsAI is an AI-powered bonsai design and sales support system for the Lingnan bonsai industry. It integrates AIGC, LLMs, MLLMs, HCI, and visualization design, covering bonsai knowledge base construction, generative model design and training, 2D/3D function development, multimodal intelligent generation, interaction design, visualization design, user study, and evaluation. It supports bonsai craftsmen, sales personnel, and customers in creative generation, design expression, interactive presentation, and marketing content production, promoting the digital and intelligent transformation of the traditional bonsai industry. 

Recognized in the “Chuang Qingchun Guangdong ‘Hundred, Thousand and Ten Thousand Project’ Rural Revitalization Youth Entrepreneurship Competition”  

Awarded “AI Rising Star” in the 2nd Baidu Wenxin Cup Entrepreneurship Competition  

Selected as one of the first representative “AI+” cases by the Guangzhou Intelligent Computing Alliance 

The Polarbrella Smart Sunshade is an innovative shading solution that integrates photovoltaic power generation with intelligent management. Equipped with an ultra-large (6㎡), highly efficient flexible perovskite module and a smart cooling system, it provides all-day shade while achieving an active temperature reduction of 5–8°C. Its built-in 1kWh energy storage module supports off-grid operation for up to 24 hours and incorporates smart features such as charging interfaces, IoT environmental sensors, and a replaceable e-ink advertising screen. Polarbrella is suited for diverse settings including urban parks, commercial plazas, and corporate campuses, meeting both municipal demands for low-carbon infrastructure and the need for distinctive experiences in commercial spaces. Multiple units can be combined to form a "Green Oasis Matrix," creating sustainably powered cool zones that redefine zero-carbon summer living through clean technology. 

Awarded in 2025 Jinko Solar Global Business and Technology Challenge 

Background

Unmanned patrol vehicles represent a significant advancement in the field of security and surveillance. These systems are designed to operate without direct human intervention, by leveraging sophisticated automation technologies. The development of Unmanned patrol vehicles addresses the growing need for efficient, continuous, and reliable surveillance in various settings. By reducing the reliance on human patrols, these vehicles enhance safety, reduce operational costs, and provide real-time data collection and analysis.

Methods and Techonology

Localization and Mapping - Use multi-modal sensor data to generate motion estimation and a 3D map of the environment.

Perception - Use deep learning to interpret the environment and recognize objects.Planning and Control - Make the robot move without collision.

Group Deliverables

All members of the group obtained extensive training from hands-on hardware and software development of an autonomous vehicle. We have practically expereinced a long-term team project involving the application of a large range of skill set and knowledge discipline.

Background

The widespread application of disposable packaging such as corrugated cardboard boxes in logistics has greatly facilitated people's production and life. However, the large amount of packaging waste generated poses a challenge to the sustainable development of human society. In response to the current high economic cost and significant environmental impact of disposable logistics packaging, this project aims to achieve packaging reduction and greening in various logistics scenarios through recyclable digital logistics turnover boxes. It also aims to improve the efficiency of the logistics system by coupling RFID and other IoT technologies with big data analysis, thereby achieving the goal of cost reduction, carbon reduction, and efficiency enhancement.

Methods and Techonology

The project's initial aspiration is to offer a suite of eco-friendly and intelligent logistics packaging solutions to the industry. It is designed systematically and encompasses a comprehensive layout of key nodes across the entire industry chain, such as material modification, equipment manufacturing, chip design, and the establishment of logistics systems. The project aims to tackle each critical node with small teams of two to three members, facilitating a chain-like synergy of innovation.

Group Deliverables

Through preliminary division of labor and field research, the team designed and initially set up a smart packaging management system based on recyclable turnover boxes for the campus food delivery scenario, reducing the use of disposable plastic bags and enabling the circular use of delivery packaging. The project won the first prize in the Campus Sustainable Smart Living Competition and was highly praised by the judges.

In the various stages of project initiation discussions, field research, plan implementation, and project roadshows, students from different academic backgrounds actively engaged in discussions, generating many sparks of inspiration. Guided by the mentors of the project team and through the concerted efforts of the students, students with a background in business and social sciences have gained a clearer understanding of how to progressively implement a project through technology and design. Students with a background in science and engineering have also learned more about considering the costs of project implementation and the potential for future commercial value in the process.