External Research Grants

This proposal presents OMNICON, a comprehensive framework for generating realistic long-term multi-human motions with environmental context. By designing novel motion representations with generative solutions, OMNICON addresses critical challenges in long-term motion generation, multi-human interactions, and motion-with-context synthesis. Designed to advance applications across animation, gaming, virtual reality, and robotics, OMNICON leverages principles from physics and spatial reasoning to produce temporally consistent, contextually adaptive, and socially coherent motion sequences.

This project, conducted in collaboration with HTX, explores the use of Generative AI (GAI) to advance scientific computing and strengthen cloud security and cybersecurity resilience. This project looks to address deep research challenges in building intelligent, domain-specific automation. This is done by leveraging LLMs for computational chemistry, cloud configuration security and developing robust defence strategies to protect AI systems for use in mission-critical settings.

The increasing realism and accessibility of AI-generated and AI-edited videos threaten public trust, information integrity, and digital security. From misinformation campaigns to identity fraud, such manipulated content can cause real-world harm. Current detection systems are limited: they often focus narrowly on facial deepfakes, lack cultural and linguistic diversity, offer little interpretability, and struggle to adapt to new manipulation techniques. Additionally, most systems emphasize passive detection, without offering mechanisms for content traceability or origin verification. This bilateral research project between Singapore Management University (SMU) and Sungkyunkwan University (SKKU) aims to address these challenges by developing an interpretable, adaptive, and globally deployable deepfake detection and protection system, tailored to the languages, dialects, and socio-cultural contexts of Singapore and South Korea. 

This research/project is supported by the National Research Foundation Singapore under the AI Singapore Programme (AISG Award No: AISG4-TC-2025-018-SGKR).

Singapore and New Zealand both use interRAI, a standardised assessment tool that supports the care of older adults. While interRAI is reliable and effective, integrating Artificial Intelligence (AI) presents a transformative opportunity to enhance healthy ageing and support older people to live longer, more independent lives. Our project brings together clinicians and researchers from the University of Otago, Singapore Management University, University of Canterbury, and University of Auckland. We will identify how to effectively integrate AI into the interRAI assessment, risk prediction, and care planning process to improve efficiency, consistency, and personalisation of care. We will achieve this with a three-pronged approach: 1. AI-assisted Assessments: By partially automating the currently manual interRAI process, we can reduce assessment time by 50% while improving accuracy. We will integrate structured health data and multimedia inputs to generate enriched assessments. 2. AI-enhanced Risk Prediction: We will develop predictive models for outcomes such as fracture risk, cognitive decline, and depression. These models will be embedded into interRAI software to support timely, targeted interventions. 3. AI-driven Personalised Care Plans: We will create dynamic, user-friendly care plans using a knowledge-based AI system enhanced by large language models. These plans will be tailored for patients, families, and clinicians, ensuring clarity and actionable guidance. With support from New Zealand’s Health NZ and ACC, and Singapore’s Agency for Integrated Care, Kwong Wai Shiu Hospital, NWC Longevity Practice, and 59 Socio-Techno Ventures, this initiative will augment existing systems to deliver scalable, cost-effective improvements to aged-care while growing our respective AI sectors. 

This research/project is supported by the National Research Foundation, Singapore under its AI Singapore Technology Challenge – Leveraging AI for Healthy Ageing (AISG Award No: AISG4-TC-2025-015-SGNZ). Any opinions, findings and conclusions or recommendations expressed in this material are those of the author(s) and do not reflect the views of National Research Foundation, Singapore.

Mild Cognitive Impairment (MCI), a precursor to dementia, offers a chance for intervention to delay dementia onset and facilitate advanced planning. In our ongoing project Sensors In-home for Elder Wellbeing (SINEW) with Sengkang General Hospital (SKH), we have the first longitudinal cohort in Asia from which clinically meaningful behavioural and digital biomarkers are captured with consumer-grade low-cost sensors installed in the homes of older adults. We have also successfully demonstrated proof-of-concept using sensors and machine learning to obtain digital phenotypes for accurate classification of MCI in a home-based setting. In this SINEW+ project, we shall build on the SINEW cohort to validate and refine our explainable predictive modelling approach for early detection of MCI. By leveraging advanced AI models and multi-modal data, we aim to implement and evaluate these solutions in real-world community settings, focusing on their cost-effectiveness and scalability to ensure broad adoption.

The objective of the proposed project is to explore, in close collaboration with a local air transport hub, the development, validation and testing of an integrated set of models, algorithms, and tools that will support the Stand Assignment Process, considering impacts on the activities and behavior of passengers within the terminals. The project will also assess the likely impacts of a new AI-based system on the range of affected stakeholders, involve managers and staff in the design process, and train them in the use and management of this technology. Similar use cases with a ride-hailing service provider are being explored.

This project targets human capital development through AI-driven learning, with a focus on both childhood and adult learners. SMU researchers will develop AI-based tutoring technologies that enhance engagement and support during self-paced learning sessions. The project includes collaboration with organizations such as Yayasan Mendaki and SMU Academy. Key objectives are to capture multi-modal learner queries – visual, verbal, and gestural – using advanced sensors, and to build AI models for interactive question answering and generation in response to such queries. Focusing initially on mathematics problems, these models will also adapt the learning content (while formally assuring the correctness of auto-generated new content) based on assessments of learners’ current levels of competency and capability. The goal is to create new AI-powered online platforms to improve learning outcomes and personalize educational experiences across diverse learner populations.

This project focuses on enabling immersive AI-assisted human-robot collaboration in dynamic industrial environments such as aviation and marine maintenance. Assistive agents deployed in robots or other wearable devices must comprehend and respond to human-issued instructions involving spatial and temporal references, adapting their behaviour in real-time. SMU researchers aim to develop lightweight, energy-efficient AI models and pervasive systems that support comprehension of such multi-modal instructions – using visual, verbal, and gestural cues– and relate them to the 3D environment captured using sensors like RGB video, LIDAR, and neuromorphic cameras. Objectives include optimizing the execution of grounding tasks (associating instructions with specific real-world objects) for moving objects using video data and developing light-weight techniques for enhanced robotic spatial reasoning and planning (e.g., navigation to retrieve specific objects). These innovations will allow robotic agents to better interpret human commands and improve task execution, ultimately enhancing safety, productivity, and the adaptability of joint human-robot collaborative work in real-world settings.

(This is additional funding to SMU with a project extension.)

This project, led by A/Prof Iris Rawtaer (Sengkang General Hospital) aims to utilise multimodal sensor networks for early detection of cognitive decline. Under this project, the SKH team will oversee the project operations, screening recruitment, psychometric evaluation, data analysis, data interpretation, reporting and answer of clinical research hypotheses. The SMU team will collaborate with SKH to provide technical expertise for this study by ensuring safe implementation and maintenance of the sensors in the homes of the participants, provide the sensor obtained data to the clinical team and apply artificial intelligence methods for predictive modelling.

(This is additional funding to SMU for Stage 1b of the project.)

State-of-the-art visual perception models in autonomous vehicles (AV) fail in the physical world when meeting adversarially designed physical objects/environmental conditions. The main reason is that they are trained with discretely-sampled samples and can hardly cover all possibilities in the real world. Although effective, existing physical attacks consider one or two physical factors and cannot simulate dynamic entities (e.g., moving cars or persons, street structures) and environment factors (e.g., weather variation and light variation) jointly. Meanwhile, most defence methods like denoising or adversarial training (AT) mainly rely on single-view or single-modal information, neglecting the multi-view cameras and different modality sensors on the AV, which contain rich complementary information. The above challenges in both attacks and defenses are caused by the lack of a continuous and unified scene representation for the AV scenarios. Motivated by the above limitations, this project firstly aims to develop a unified AV scene representation based on the neural implicit representation to generate realistic new scenes. With this representation, we will develop extensive physical attacks, multi-view & multi-modal defenses, as well as a more complete evaluation framework. Specifically, the project will build a unified physical attack framework against AV perception models, which can adversarially optimize the physical-related parameters and generate more threatening examples that could happen in the real world. Furthermore, the project will build the multi-view and multi-modal defensive methods including a data reconstruction framework to reconstruct clean inputs and a novel ‘adversarial training’ method, i.e., adversarial repairing that enhances the robustness of the deep models with guidance of collected adversarial scenes. Finally, a robust-oriented explainable method will be developed to understand the behaviors of visual perception models under physical adversarial attacks and robustness enhancement.